essays on growth and human capital : an analysis of education policy
TRANSCRIPT
AIX-MARSEILLE UNIVERSITÉ
FACULTÉ D’ECONOMIE ET DE GESTION
École Doctorale de Sciences Economiques et de Gestion d’Aix-Marseille No 372
Année 2014 Numéro attribué par la bibliothèque :
a a a a a a a a a a
Thèse pour le Doctorat ès Sciences Economiques
Présentée et soutenue publiquement le 5 Decembre 2014 par
MARION DAVIN
Essays on growth and human capital : an
analysis of education policy
Directeurs de thèse :
Karine Gente - Maître de Conférence à l’Université d’Aix-Marseille, AMSE
Carine Nourry - Professeur à l’Université d’Aix-Marseille & Membre de l’IUF,
AMSE
Jury :
Andreas Irmen, Rapporteur - Professeur à l’Université du Luxembourg
Katheline Shubert, Rapporteur - Professeur à l’Université Paris 1
Panthéon-Sorbonne
Alain Venditti, Examinateur - Directeur de recherche CNRS, AMSE
L’université d’Aix-Marseille n’entend ni approuver, ni désapprouver les opinions par-ticulières du candidat : Ces opinions doivent être considérées comme propres à leur auteur.
iii
Résumé
Cette thèse se compose de quatre essais théoriques, portant sur le capital humain et
la croissance. L’objectif est de proposer de nouvelles approches afin de mieux identifier
l’impact des politiques éducatives. Plus précisément, dans les chapitres un à trois, nos ana-
lyses sont menées dans un cadre à deux secteurs, dans la mesure où les différents secteurs
qui composent une économie ne sont pas influencés de la même façon par l’éducation.
Dans le quatrième chapitre nous nous intéressons aux enjeux politiques liés à la gestion
des problèmes environnementaux en considérant le lien entre éducation et environnement.
Le premier chapitre se focalise sur l’investissement optimal en capital humain et phy-
sique. En privilégiant une approche utilitariste, qui consiste à prendre en considération les
sentiments altruistes des agents dans la fonction de bien-être social, nous montrons que
les différences d’intensité en facteur entre les secteurs sont déterminantes : en fonction des
préférences des agents et du taux d’actualisation social du planificateur, une variation de
ces intensités modifie l’orientation de la politique optimale en faveur du capital physique
ou de l’éducation.
Le deuxième chapitre examine le lien entre éducation publique et croissance écono-
mique. Lorsque l’éducation publique est financée par des taxes sectorielles, la politique
éducative qui maximise le taux de croissance diffère de celle dictée par un financement
uni-sectoriel. Deux mécanismes expliquent cela. Tout d’abord, les préférences des agents
en capital humain, en service et en épargne affectent la relation croissance-éducation pu-
blique dès lors que des taxes sectorielles sont considérées. Ensuite, ce type de financement
crée une distorsion, en modifiant le prix relatif de l’éducation. Nous montrons ainsi qu’un
financement de l’éducation publique par des taxes sectorielles peut conduire à un taux de
croissance de long-terme plus élevé qu’une taxe sur la production agrégée.
Le troisième chapitre porte sur les effets de court et de long-terme de l’intégration
économique. Nous soulignons que l’intégration peut-être couteuse à court-terme mais bé-
néfique à long-terme, apportant une évidence tangible sur le fait que l’analyse de long-
terme n’est pas suffisante. De plus, nous montrons que les différences de productivité entre
le secteur échangeable domestique et étranger sont les principaux déterminants des effets
de l’intégration. D’un point de vue politique, ce chapitre préconise de soutenir les actions
visant à favoriser les externalités transfrontalières en éducation, à travers la mobilité des
étudiants par exemple.
Le quatrième chapitre examine la relation entre politique environnementale et crois-
sance lorsque les préférences vertes sont déterminées de façon endogène par le niveau
d’éducation d’un agent et la pollution du pays. Nous schématisons une politique environ-
nementale avec un revenu recycling : le gouvernement peut implémenter une taxe sur les
activités polluantes et allouer le revenu de cette taxe à de la maintenance publique ou à
du soutient indirect à l’environnement en subventionnant l’éducation. Cette politique peut
permettre d’améliorer le taux de croissance de long-terme et d’éliminer les oscillations
amorties, source d’inégalités intergénérationnelles, induites par les préférences endogènes.
Mots clefs : Altruisme paternaliste, Conscience environnementale, Croissance endo-
gène, Education, Intensité factorielle, Intégration économique, Maintenance environne-
mentale, Modèle à deux secteurs, Optimum social.
v
Abstract
This dissertation consists of four essays on human capital and growth. It aims at pro-
posing approaches to better understand the influence of education policy. Specifically, we
take into account sectoral properties, since education does not affect each sector in the
same way. We also deal with the link between education and the environment, to address
environmental challenges that are one of the major political issues.
The first chapter focuses on the optimal investment in human and physical capital.
Considering an optimal view point, we highlight that the definition of the social planner
welfare function is not innocuous to determine the optimal accumulation of factors. By
considering a Utilitarian view, we prove that relative factor intensity between sectors dras-
tically shapes the welfare analysis : two identical laissez-faire economies with different
sectoral capital shares may generate physical capital excess or scarcity, with respect to the
optimum. When the social planner mainly values future generations, changes occurring in
the investment sector prevail : more resources have to be devote to education when invest-
ment sector becomes more human capital intensive.
The second chapter examines the interplay between public education expenditure and
economic growth. When public education is financed by sectoral taxes, the education po-
licy maximizing the growth rate differs from that obtained by the standard unisectoral tax.
The reasons for this are twofold. First, because agents’ preferences for services, human
capital and savings become a major determinant of the relationship between growth and
public education expenditure. Second, because education spending is a service and hence
sectoral taxation creates a distortion by affecting its relative price. Finally, we reveal that
a sectoral tax may perform better than a standard aggregate production tax in terms of
long-term growth.
vi
The chapter three deals with short and long-term effects of economic integration. We
reveal that integration may be growth damaging in the short run, while it turns out to be
growth improving in the long run. This result provides more tangible evidences that eva-
luate only the long-term impact of economic integration is insufficient. Moreover, agents’
preferences for services and traded-TFPs disparities across countries influence the overall
impact of economic integration. From a policy perspective, this chapter emphasizes that
policy makers should pursue their efforts to promote student mobility, as the presence of
cross border externalities is beneficial in the context of economic integration.
The chapter four examines the relationship between environmental policy and growth
when green preferences are endogenously determined by education and pollution. The go-
vernment can implement a tax on pollution and recycle the revenue in public pollution aba-
tement and/or education subsidy (influencing green behaviors). When agent’s preferences
for the environment is highly sensitive to human capital and environmental changes, the
economy can converge to a balanced growth path equilibrium with damped oscillations.
We show that an environmental policy can remove these oscillations, source of intergene-
rational inequalities, and enhance the long-term growth rate. For that, the revenue from a
tighter tax has to be well allocated between education and direct environmental protection.
This paper concludes in favor of the implementation of policies mix to answer environ-
mental and economic issues.
Keywords : Economic integration, Education, Endogenous growth, Environmental
awareness, Environmental maintenance, Factor intensity differential, Paternalistic altruism,
Social optimum, Two-sector model.
vii
Remerciement
J’ai bénéficié d’un environnement riche pendant ces quatre années de thèse et je vais
tenter de remercier, dans ces quelques lignes, l’ensemble des personnes qui ont contribué,
de près ou de loin, à la réalisation de mon travail de recherche.
Il convient en premier lieu d’adresser mes sincères remerciements à mes directrices de
thèse, Karine Gente et Carine Nourry. J’ai d’abord découvert leur qualité en tant qu’ensei-
gnante pendant mon master et l’encadrement qu’elles m’ont fourni pendant mon doctorat
m’a permis d’apprécier leur qualité humaine et scientifique. Je leur adresse ma profonde
gratitude pour leurs conseils avisés et leur implication, qui m’ont permis de mener à bien
mes travaux. Je les remercie pour le temps précieux et la confiance qu’elles m’ont ac-
cordée ainsi que pour leur soutien général et leur gentillesse. J’ai énormément appris en
travaillant à leurs côtés.
Je tiens à remercier chaleureusement Andreas Irmen et Katheline Schubert qui ont
accepté d’être les rapporteurs de cette thèse. Leurs précieux conseils donnés pour ma pré-
soutenance ont largement contribué à l’amélioration des différents chapitres de cette thèse.
Un grand merci également à Alain Venditti, pour avoir accepté de participer à mon jury
de thèse bien sûr, mais également pour l’ensemble des remarques constructives qu’il m’a
fourni pendant ma pré-soutenance et tout au long de ma thèse, notamment lors de mes
présentations au groupe de travail macro.
Je suis extrêmement reconnaissante à Thomas Seegmuler qui, en tant que spécialiste
viii
de l’économie de l’environnement, m’a largement encadré et supervisé pour le quatrième
chapitre de cette thèse. Je le remercie sincèrement pour ses conseils, son soutien et pour le
temps non négligeable qu’il m’a accordé.
J’ai été heureuse d’avoir l’opportunité de discuter de mes travaux avec des chercheurs
invités au GREQAM. Je remercie, dans l’ordre alphabétique, Michael Devereux, Mouez
Fohda, Omar Licandro, Oded Galor, Olivier Loisel et Xavier Raurich pour leurs sugges-
tions.
J’adresse mes remerciements à la direction du GREQAM et de l’AMSE, pour les fonds
qu’ils m’ont accordé pour mes conférences et mon séjour de recherche à l’Université de
LAVAL. Je tiens d’ailleurs à remercier Michel Normandin ainsi que Sylvain Dessy du
CIRPEE, qui m’ont permis d’effectuer ce séjour enrichissant. Je tiens aussi à remercier
l’Ecole doctorale 372 d’avoir participé au financement de mes déplacements. Merci égale-
ment à toute l’équipe administrative du GREQAM qui nous facilite le travail au quotidien :
Bernadette, Elizabeth, Isabelle et Mathilde, et aussi Agnès, Aziza, Carole, Corinne, Gre-
gory et Yves.
Je remercie l’ensemble de mes collègues doctorants, actuels et anciens, qui m’ont ac-
compagnée tout au long de cette thèse et qui ont grandement enrichi cette aventure. Dans
l’ordre alphabétique, Antoine Bonleu, Antoine Le Riche, Anwar, Bilel, Camila, Clémen-
tine, Cyril, Daria, Emma, Florent, Joao, Kadija, Kalila, Lise, Manel, Maty, Nick (merci
pour la relecture de l’intro du troisième chapitre), Nariné, Nicolas Abad, Nicolas Caudal,
Pauline, Régis, Thomas, Vincent, Vivien, Waqar, Zakaria. J’ai une pensée particulière pour
les occupants actuels et passés du bureau 203 de la Vieille Charité, dans lequel la convivia-
lité ambiante (en témoignent les nombreuses bougies soufflées dans le bureau) durant ces
quatre années a rendu cette aventure si particulière. Un grand merci à Anastasia et Martha
ix
pour leur formidable gentillesse et leur soutien mais surtout pour leur amitié sincère. Je
remercie également Natasha qui m’a très souvent soutenue et réconfortée.
Ces remerciements seraient incomplets si je ne mentionnais pas Gilles et Karine, ren-
contrés dans l’amphi A de Ferry en 2005 (dans un certain cours d’histoire des faits et de la
pensée si je me souviens bien...) et qui ont traversé cette étape du doctorat avec moi. Neuf
longues années d’étude ensemble mais surtout neuf années d’amitié. Je vous remercie
pour tant de choses : pour m’avoir épaulée, pour m’avoir accueillie si souvent chez vous
à Aix, pour les week-ends à Ollioules, pour les déjeuners au château. . . Votre présence a
grandement contribué à l’avancement de cette thèse. Un remerciement supplémentaire à
Karine qui en plus de ses « fonctions » d’amie, de confidente, de témoin et de correctrice
est désormais ma co-auteur. Merci de m’avoir initiée à l’économie de l’environnement, ce
fut un plaisir de coécrire mon quatrième chapitre de thèse avec toi.
Enfin, je remercie sincèrement ma famille à qui je dédie cette thèse. Mes sœurs, Marie-
Pierre, Marianne et Marie-Octavie, ainsi que mon frère Matthieu, qui m’ont toujours fait
confiance et qui ont su me soutenir en toutes circonstances depuis le début de mes études.
Les occasions sont rares de pouvoir remercier ses proches par écrit, j’en profite donc pour
vous dire combien je suis heureuse et reconnaissante de vous avoir à mes côtés. Une men-
tion spéciale pour la petite dernière, Marie Oc, à qui je souhaite beaucoup de réussite pour
ses études futures. Mes derniers remerciements s’adressent évidemment à Matthias, mon
mari, qui a subi mon stress, mes doutes et mes sautes d’humeur pendant mes longues
études. Merci pour ta compréhension, ta présence et ton soutien sans faille.
x
Avertissement
Les différents chapitres de cette thèse sont issus d’article de recherche rédigés en anglais
et dont la structure est autonome. Ceci y explique la présence de terme "paper" ou "article"
ainsi que l’éventuelle répétition de certaines informations.
Notice
The chapters of this dissertation are self-containing research articles. Consequently, terms
"paper" and "article" are frequently used. This also explains that some informations are
given in multiple places of the thesis.
xii
Table of content
Introduction 1
1 Optimal human and physical capital accumulation 36
A Social optimum in an OLG model with paternalistic altruism 36
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.2 Social optimum and paternalistic altruism . . . . . . . . . . . . . . . . . 38
1.3 Optimal growth and capital accumulation . . . . . . . . . . . . . . . . . 43
B Should a country invest more in human or physical capital ? A two-sector
endogenous growth approach 49
1.4 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
1.5 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
1.5.1 The production structure . . . . . . . . . . . . . . . . . . . . . . 52
1.5.2 Household’s behavior . . . . . . . . . . . . . . . . . . . . . . . . 54
1.5.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
1.6 The social planner’s problem . . . . . . . . . . . . . . . . . . . . . . . . 58
1.7 Laissez-faire and the social optimum . . . . . . . . . . . . . . . . . . . . 64
1.8 Policy implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
1.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
1.10 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
1.10.1 Proof of Lemma 1 . . . . . . . . . . . . . . . . . . . . . . . . . 73
1.10.2 Proof of Lemma 2 . . . . . . . . . . . . . . . . . . . . . . . . . 75
xiii
1.10.3 Proof of Proposition 4 . . . . . . . . . . . . . . . . . . . . . . . 78
1.10.4 Proof of Proposition 5 . . . . . . . . . . . . . . . . . . . . . . . 79
1.10.5 Proof of Proposition 6 . . . . . . . . . . . . . . . . . . . . . . . 80
1.10.6 Proof of Corollary 2 . . . . . . . . . . . . . . . . . . . . . . . . 81
2 Public education spending, sectoral taxation, and growth 82
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.2.1 Production technologies . . . . . . . . . . . . . . . . . . . . . . 85
2.2.2 Government . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.2.3 Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
2.2.4 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.3 Public education funding and long-term growth rate . . . . . . . . . . . 92
2.3.1 Public education financed by a tax on aggregate production . . . . 92
2.3.2 Public education financed by a tax on manufacturing output . . . 94
2.3.3 Public education financed by a tax on services . . . . . . . . . . 97
2.4 Sectoral tax versus aggregate output tax . . . . . . . . . . . . . . . . . . 100
2.5 Extension : factor intensity differential between sectors . . . . . . . . . . 102
2.6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
2.7 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
2.7.1 Proof of Lemma 4 . . . . . . . . . . . . . . . . . . . . . . . . . 104
2.7.2 Proof of Lemma 5, 6 and 7 . . . . . . . . . . . . . . . . . . . . . 105
2.7.3 Proof of Proposition 12 . . . . . . . . . . . . . . . . . . . . . . . 106
3 Short-and long-term growth effects of integration in two-sector
economies with non-tradable goods 108
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.2 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.2.1 Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.2.2 Consumption, savings and children’s education . . . . . . . . . . 114
3.2.3 Cross-border external effects in human capital . . . . . . . . . . . 118
xiv
3.2.4 The non-tradable market clearing condition . . . . . . . . . . . . 119
3.3 Autarky . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
3.4 Economic integration and growth . . . . . . . . . . . . . . . . . . . . . . 124
3.4.1 International environment . . . . . . . . . . . . . . . . . . . . . 125
3.4.2 Steady state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.4.3 Dynamics and short-term implications . . . . . . . . . . . . . . . 128
3.4.4 Long-term integration benefits . . . . . . . . . . . . . . . . . . . 133
3.5 A numerical example . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
3.7 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
3.7.1 Non-tradable market equilibrium : Proof of Lemma 8 . . . . . . . 143
3.7.2 Proof of Lemma 9 . . . . . . . . . . . . . . . . . . . . . . . . . 144
3.7.3 Proof of Proposition 15 . . . . . . . . . . . . . . . . . . . . . . . 145
3.7.4 Proof of Lemma 10 . . . . . . . . . . . . . . . . . . . . . . . . . 146
3.7.5 Proof of Lemma 11 . . . . . . . . . . . . . . . . . . . . . . . . . 147
3.7.6 Proof of Proposition 17 . . . . . . . . . . . . . . . . . . . . . . . 148
3.7.7 Proof of Proposition 18 . . . . . . . . . . . . . . . . . . . . . . . 150
3.7.8 Proof of Corollary 5 . . . . . . . . . . . . . . . . . . . . . . . . 151
4 Environmental policy and growth in a model with endogenous
environmental awareness 153
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
4.2 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.2.1 Consumer’s behavior . . . . . . . . . . . . . . . . . . . . . . . . 158
4.2.2 Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.2.3 The government . . . . . . . . . . . . . . . . . . . . . . . . . . 163
4.2.4 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
4.3 Balanced growth path and transitional dynamics . . . . . . . . . . . . . . 167
4.4 Environmental policy implications . . . . . . . . . . . . . . . . . . . . . 171
4.4.1 The short-term effect of environmental tax . . . . . . . . . . . . 172
4.4.2 The long-term effect of environmental tax . . . . . . . . . . . . . 174
xv
4.4.3 How the government policy can improve the short-and long-term
situations ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
4.6 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
4.6.1 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
4.6.2 Proof of Proposition 19 . . . . . . . . . . . . . . . . . . . . . . . 185
4.6.3 Proof of Proposition 20 . . . . . . . . . . . . . . . . . . . . . . . 188
4.6.4 Proof of Proposition 21 . . . . . . . . . . . . . . . . . . . . . . . 191
4.6.5 Proof of Proposition 22 . . . . . . . . . . . . . . . . . . . . . . . 192
General conclusion 195
References 201
xvi
List of tables
1.1 Parameters Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.2 Calibration results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1 Long-term growth impact of integration . . . . . . . . . . . . . . . . . . 135
3.2 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.3 Long-term impact of integration on growth . . . . . . . . . . . . . . . . 139
xvii
List of figures
1.1 Gap between the optimal and the laissez-faire physical to human ratio . . 67
3.2 Global Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.3 Growth impact of economic integration at t = 1 . . . . . . . . . . . . . . 141
4.4 Dynamics when N > N . . . . . . . . . . . . . . . . . . . . . . . . . . 170
4.5 Short-and long-term implications of a tighter tax . . . . . . . . . . . . . . 179
4.6 Function J at given τ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
xviii
Introduction générale
0.1 Croissance économique et éducation
0.1.1 Capital humain et croissance endogène
Le rôle de l’éducation dans le domaine économique a été mis en évidence par la théorie
du capital humain, dont les fondements proviennent des travaux de Schultz (1961). Schultz
donne une dimension qualitative au facteur travail, en soulignant que les travailleurs dis-
posent d’un capital humain qui influence la productivité. Il identifie les dépenses de for-
mation, d’infrastructure ou dans le système éducatif, comme des sources d’amélioration
du capital humain. L’analyse de Becker (1964) prolonge les travaux de Schultz (1961), en
s’intéressant aux comportements des agents et à la façon dont le capital humain est accu-
mulé. Selon lui, l’investissement en capital humain résulte d’un calcul coût-avantage de la
part des agents économiques rationnels. Il affine ainsi le concept de capital humain, qu’il
définit comme un stock de connaissance qu’un individu acquiert tout au long de sa vie par
le biais d’investissement et qui est susceptible de lui procurer un revenu monétaire futur.
Ces théories donnent à l’éducation sa dimension centrale. Les dépenses en éducation sont
identifiées comme des investissements qui contribuent à l’accroissement de la qualité de
la main d’œuvre et sont donc des facteurs permettant d’alimenter l’acquisition de capital
humain.
Dans le prolongement de la théorie du capital humain, de nombreuses études se sont
focalisées sur une approche microéconomique de l’éducation, en déterminant le rende-
ment de l’éducation en fonction de son impact sur le salaire (Mincer, 1974) ou le chômage
1
Introduction générale
(Mincer, 1991). Ces analyses fournissent des informations essentielles pour appréhender
les bénéfices des dépenses éducatives, néanmoins elles ne prennent pas en considération
les externalités multiples associées à un niveau de qualification élevé. Parmi les plus évi-
dentes, nous pouvons citer les externalités entre individus au sein de la sphère productive :
les travailleurs les plus éduqués et les plus productifs tendent à stimuler la productivité des
travailleurs moins qualifiés. À un niveau plus agrégé, l’éducation favorise également le
progrès technique et sa diffusion. Des externalités entre générations sont également clai-
rement identifiées. En effet, les parents jouent un rôle déterminant dans l’acquisition du
savoir des enfants : le niveau de connaissance d’un individu dépend de l’éducation de ses
parents, ce qui tend à freiner la mobilité intergénérationnelle, définie comme le change-
ment de position sociale d’un agent par rapport à celle de ses parents. Une littérature plus
récente a mis en évidence que les externalités associées à l’éducation s’étendaient bien au-
delà de la sphère purement économique. Par exemple, le quality-quantity trade-off entre
éducation et fertilité, mis en évidence par De la Croix and Doepke (2003), montre que
l’accroissement de l’éducation favorise une baisse de la fertilité, nécessaire à la transition
démographique. Les dépenses éducatives tendent également à favoriser l’accès à la santé.
Comme le souligne Mirowsky and Ross (1998), celles-ci permettent, à travers un effet
revenu, d’améliorer le mode de vie et ont tendance à encourager l’adoption de comporte-
ments plus sains. Enfin, l’éducation est également perçue comme un facteur favorisant la
sensibilité aux problèmes environnementaux (Prieur and Bréchet, 2013).
Au regard de ces effets externes, l’approche macroéconomique des rendements de
l’éducation apparait cruciale. Dans cette perspective, le modèle de croissance endogène
développé par Lucas (1988) est le point de départ d’une vaste littérature portant sur le
lien entre croissance économique et investissement en éducation. Lucas (1988) s’efforce
2
Introduction générale
d’expliquer l’effet positif de l’éducation, et plus généralement de l’accumulation de capi-
tal humain, sur la croissance économique. Il développe un cadre théorique où le niveau
de production d’une économie dépend du stock de capital humain. Les agents maximisent
leur utilité inter-temporelle en allouant leur temps, non dédié au loisir, entre la production
et la formation. Selon lui, le capital humain individuel accroit la productivité des agents
(effet interne) et le niveau moyen de capital humain dans l’économie accroit la productivité
globale des facteurs (effet externe). L’intérêt du modèle de Lucas réside dans la formalisa-
tion d’une loi d’accumulation spécifique au capital humain, qui prend en compte l’effort
de formation des agents. 1 L’hypothèse d’une production de capital humain à rendement
constant assure le caractère auto-entretenu de la croissance. La modélisation proposée par
Lucas suggère qu’à long terme, une croissance soutenue n’est envisageable que si le capi-
tal humain augmente de manière permanente. Un accroissement du niveau de qualification
de la population active est donc nécessaire pour soutenir la croissance économique.
Dans la continuité de Lucas (1988), de nombreux travaux théoriques ont considéré le
rôle conducteur de l’éducation sur la croissance. Ces études ont donné lieu à d’abondantes
investigations empiriques, visant à vérifier la validité du lien positif entre croissance et ca-
pital humain. L’article pionnier de Mankiw et al. (1992), dans lequel le capital humain est
un facteur de production au même titre que le capital physique, conclut à un effet positif de
l’éducation sur le niveau de croissance. Plus précisément, il montre que le taux de scolari-
sation des agents entre 12 et 17 ans a un effet positif et significatif sur le PIB par tête pour
les pays de l’OCDE. Néanmoins, les études suivantes ont donné lieu à des conclusions
contradictoires remettant en question la relation entre capital humain et croissance prédite
par l’analyse théorique. Selon Benhabib and Spiegel (1994), c’est le niveau de capital hu-
1. Uzawa (1965) est le premier à introduire explicitement une fonction d’accumulation de capital humainmais il ignore l’effort des agents dans la création de connaissance.
3
Introduction générale
main et non son taux de croissance (et donc son accumulation), qui améliore la taux de
croissance du PIB par tête. On peut également citer le papier de Prichett (2001) qui ne
trouve pas d’association entre l’amélioration du capital humain, attribuable à la hausse de
la qualification des travailleurs, et le taux de croissance de la production par tête. Une revue
de la littérature de Krueger and Lindahl (2001) remet en cause ce scepticisme sur le rôle
de la formation dans le processus de croissance, en révélant des limites dans les méthodes
d’estimations utilisées par ces auteurs. Notamment, Krueger and Lindahl (2001) met en
évidence des erreurs de mesure relatives au niveau d’éduction. Celles-ci tendent à sous es-
timer l’effet de l’accroissement du niveau d’éducation sur la croissance économique. Des
contributions plus récentes ont d’ailleurs confirmé ce point de vue. Les études empiriques
recherchant à estimer l’effet du capital humain sur la croissance font face à plusieurs dif-
ficultés. Tout d’abord, comme le souligne De la Fuente and Doménech (2006), la faible
qualité des données sur l’éducation qui sont utilisées peut conduire à des résultats contrin-
tuitifs. Ensuite, ces analyses se heurtent au concept de capital humain. Cohen and Soto
(2007) révèlent que les variables éducatives utilisées comme proxy du capital humain dans
ces analyses empiriques sont imprécises. En utilisant des bases de données plus détaillées
et des méthodes économétriques plus sophistiquées, De la Fuente and Doménech (2006)
et Cohen and Soto (2007) montrent que les variables d’éducation dans les régressions de
croissance ont un effet positif et significatif. Ils confirment ainsi, empiriquement, le rôle
déterminant de l’éducation.
4
Introduction générale
0.1.2 L’apport des modèles à générations imbriquées pour traiter des questions écono-
miques liées à l’éducation
La persistance intergénérationnelle des niveaux d’éducation observée dans de nom-
breux pays suggère que les liens familiaux et les transferts entre générations jouent un
rôle déterminant dans le processus de formation du capital humain. Ces faits ont donné
lieu à de nombreuses études sur la mobilité intergénérationnelle, qui visent à identifier
les facteurs favorisant la reproduction sociale. Une étude empirique de Oreopoulos et al.
(2006), portant sur les Etats-Unis, montre que l’éducation des parents a un effet significatif
et positif sur l’accumulation de capital humain des enfants. Une augmentation d’une an-
née de formation des parents réduit la probabilité qu’un enfant redouble de 2 à 7 points de
pourcentage. Cette forte corrélation entre le statut social des parents et des enfants résulte
en partie de transferts intergénérationnels privés, monétaires ou non-monétaires, affectant
la formation de capital humain. Les transferts monétaires regroupent les dépenses pécu-
niaires faites par les parents pour financer les études de leurs enfants. De tels transferts sont
généralement volontaires et résultent d’un choix des parents de contribuer à la formation
de leurs enfants. Ils sont complétés par différents transferts non-monétaires. C’est le cas
du temps accordé au soutien scolaire par exemple. Néanmoins, l’influence des parents ne
se limite pas à ces transferts mais résulte également de spillovers que l’agent ne contrôle
pas. La famille est un système social dans lequel l’agent développe ses compétences cog-
nitives et sociales. Par conséquent, le niveau de connaissances d’un individu est influencé
par son environnement familial, qui lui-même est déterminé par le niveau de qualification
de ses parents. Une part du capital humain est ainsi transmise de façon indirecte entre
générations. Une façon simple et souvent retenue pour modéliser ce type de transmission
est de supposer que la fonction d’accumulation de capital humain d’un agent dépend du
5
Introduction générale
stock de capital humain de la génération précédente. Cette formalisation rend possible une
croissance entretenue du capital humain.
Les modèles à agents représentatifs, tel que celui proposé par Lucas (1988), permettent
d’étudier l’affectation temporelle des ressources et son impact sur l’accumulation des
connaissances. Toutefois, ils ne fournissent pas un cadre adéquat pour prendre en compte
les différents liens intergénérationnels. Les modèles à générations imbriquées, introduit
par Allais (1947) puis développés par Samuleson (1958) et Diamond (1965), permettent
de surpasser cette limite. Cette modélisation permet de scinder le cycle de vie fini d’un
agent en au moins deux périodes. L’agent se consacre à ses études durant sa première
période de vie et retire les bénéfices de son éducation aux périodes suivantes. L’article
pionnier de Glomm and Ravikumar (1992) a mis en lumière l’intérêt de la structure à gé-
nérations imbriquées pour traiter de questions économiques liées à l’éducation, et a conduit
de nombreux auteurs à privilégier cette approche. Dans leur modèle, Glomm and Raviku-
mar (1992) justifient l’implication des parents dans la formation de leur enfant par une
forme d’altruisme familial, selon lequel les parents retirent satisfaction de l’éducation de
leurs enfants. Cet argument trouve ses fondements dans l’analyse de Becker (1991) :
“Yet many economists dispute that altruism is important in families, even though these
same economists often deny themselves in order to accumulate gifts and bequests for their
children. Moreover, parental love, especially mother love, has been recognized since bibli-
cal times.” Becker (1991), p. 9.
Comme le suggère Cremer and Pestieu (2006) plusieurs motivations, et donc plusieurs
formes d’altruisme, expliquent l’effort couteux d’éducation entrepris par les parents. Ils
identifient trois raisons centrales : un motif d’échange, les parents anticipent une récipro-
cité de la part de leurs enfants, un altruisme “limité” selon lequel les parents valorisent le
6
Introduction générale
capital humain de leurs enfants, ou leur contribution à ce capital 2, et enfin un altruisme
pur à la Barro, selon lequel les parents valorisent l’utilité inter-temporelle de leurs enfants,
et donc le bien-être des générations futures. Dans la majorité des études, c’est la formali-
sation d’un altruisme “limitée” qui est privilégié pour justifier des dépenses en éducation.
L’altruisme pur à la Barro trouve moins de validité empirique.
Ainsi, dans Glomm and Ravikumar (1992), le capital humain s’accumule selon le pro-
cessus suivant :
ht+1 = A(1− nt)βeγt h
δt
Le stock de capital humain d’un adulte né en t dépend de trois facteurs. Du temps que
l’agent décide d’allouer au loisir pendant son enfance, nt, des dépenses éducatives privées
en unité de biens faites par ses parents, et, et enfin du stock de capital humain de ses
parents ht. Les termes β, γ et δ capturent respectivement l’élasticité de ht+1 à chacun
de ces facteurs. Les choix de l’agent résultent de la maximisation de leur utilité inter-
temporelle :
lnnt + ln ct+1 + ln et
sous contrainte de budget et d’accumulation du capital. L’agent valorise ici sa contribution
financière au capital humain de son enfant. Cette formalisation est fréquemment étendue
en supposant que l’individu valorise plutôt le stock global de capital humain de son en-
fant. Les parents influencent l’acquisition de connaissance de leurs enfants par deux fac-
teurs dans ce modèle : par un transfert de ressources, motivé par l’altruisme, et par une
transmission indirecte à travers le processus d’accumulation du capital humain. La modé-
lisation proposée par Glomm and Ravikumar (1992) a été étendue, notamment par De la
2. Nous parlerons d’altruisme paternaliste lorsque l’agent valorise le capital humain de son enfant etd’un altruisme de type ““joy of giving” lorsqu’il valorise directement sa contribution au capital humain.
7
Introduction générale
Croix and Monfort (2000), pour introduire du capital physique, dont l’accumulation est
alimentée par de l’épargne privée. La prise en compte de ces deux formes de capital est
largement souhaitable mais l’une des limites de ce modèle réside dans l’hypothèse que le
même secteur produit un bien composite qui va servir à consommer et à investir en capi-
tal humain et physique. De plus, il suppose implicitement que l’ensemble des secteurs de
l’économie utilise ces deux facteurs de production dans des proportions identiques. Cette
restriction sera levée dans les chapitres un à trois de la thèse.
0.1.3 Capital physique et capital humain : l’intérêt d’une approche multisectorielle
“The one-sector production model, while an essential analytical tool of aggregate eco-
nomics, is inevitably limited as a framework for analyzing the full economy-wide impacts
of policy shocks and structural changes [...] one needs to augment the basic model to in-
troduce a second, or perhaps even more, sectors.” Turnovsky (2009), p.104.
Cet argument est particulièrement fondé lorsque l’on modélise la structure productive
d’une économie avec deux facteurs de production dont les caractéristiques diffèrent fonda-
mentalement. Comme nous l’enseignent les différentes théories énoncées précédemment,
le capital humain regroupe des aptitudes incorporées à une personne et revêt ainsi un carac-
tère inaliénable. Il n’est pas accumulé par le même processus et ne résulte pas des mêmes
investissements que le capital physique. Dans la forme la plus générale, le capital humain
résulte de dépenses éducatives tandis que le capital physique résulte de choix d’épargne,
qui, à l’équilibre, déterminent le montant de l’investissement alloué au capital physique.
Chaque capital a donc des spécificités propres, de sorte que ces facteurs ne sont pas
utilisés de la même façon dans le processus de production. La structure productive de
8
Introduction générale
l’économie se décompose en firmes dont l’intensité en capital physique et humain diffère.
Certaines activités, comme les services, utilisent intensément du capital humain tandis
que les secteurs de l’industrie, du transport ou de la construction sont intensifs en capital
physique. Une distinction entre secteur intensif en capital humain et physique semble donc
importante pour mieux appréhender les effets de l’éducation sur la croissance économique.
Le modèle Uzawa-Lucas, largement utilisé dans les modèles de croissance optimale à du-
rée de vie infinie, offre une formalisation intéressante puisque qu’il distingue deux secteurs
et deux facteurs de production. Un premier secteur produit un bien composite, qui sert pour
la consommation et le capital physique, et un deuxième secteur sert à la formation du ca-
pital humain. Dans la forme initiale du modèle, le capital physique est utilisé uniquement
dans le premier secteur, ce qui ne permet pas de prendre en compte les différences d’inten-
sité capitalistique entre les secteurs. Une modélisation plus générale, où les deux secteurs
utilisent les deux facteurs de production, a été proposée par Rebelo (1991) et permet de
supposer que la formation de capital humain requiert du capital physique (librairies, maté-
riels informatiques, laboratoires etc.) et du capital humain. Cette formalisation a été large-
ment utilisée dans la littérature sur la croissance endogène. Des auteurs se sont d’ailleurs
attelés à étudier la dynamique de transition des sentiers de croissance équilibrée et ont
souligné l’influence du différentiel d’intensité factorielle entre les secteurs (par exemple
Bond et al., 1996; 2003, Hu et al., 2009). D’un point de vue empirique, l’existence d’un
différentiel d’intensité en facteurs entre les secteurs est confirmée par des études qui ont
évalué l’intensité en travail et en capital de différents secteurs (Herrendorf and Valentinyi,
2008, Zuelta and Young, 2012). Par ailleurs, de récents travaux ont également souligné
que ces intensités tendent à évoluer durant les phases de croissance. Takahashi et.al (2012)
agrègent l’économie en deux secteurs, consommation et investissement, et montrent que
9
Introduction générale
l’intensité capitalistique diffère entre ces deux secteurs pour plusieurs pays de l’OCDE.
De plus, ils mettent en avant un renversement des intensités factorielles relatives entre les
secteurs au Japon. Après le choc pétrolier de 1973, le secteur d’investissement, qui était
auparavant intensif en capital physique, devient intensif en travail. Ce phénomène révèle
l’importance des spécificités sectorielles et nous amène à nous questionner sur les inci-
dences qu’elles peuvent avoir sur l’accumulation de capital humain.
Les questions de politiques économiques relatives à l’éducation, et les liens intergéné-
rationnels qui s’y rattachent, ne sont pas considérées dans ce cadre particulier. Les fonde-
ments des modèles à générations imbriqués à deux secteurs ont été présentés dans le début
des années 90 avec le modèle de Galor (1992a). Il considère que le capital physique et le
travail sont utilisés par un secteur de consommation et un secteur d’investissement avec
des intensités différentes. La qualité des travailleurs n’est pas prise en compte et le capi-
tal physique est le seul facteur de production accumulé de façon endogène. L’utilisation
d’une structure à deux secteurs avec capital physique et capital humain endogène est peu
répandue dans le cadre à générations imbriquées. A notre connaissance, le seul modèle
proposé est celui de Erosa et al. (2010). Ils distinguent un secteur manufacturé et un sec-
teur de service qui utilisent tous deux du capital physique et humain. Le bien manufacturé
peut être consommé ou investi en capital physique et les services peuvent être consommés
ou investis en capital humain. Le secteur des services produit ainsi un input, les dépenses
éducatives en services, qui va intervenir dans la fonction de production du capital humain.
Erosa et al. (2010) prennent également en considération les différences de productivités
globales des facteurs entre les secteurs. Nous verrons que, dans un contexte international,
les différences de productivité sectorielle entre les pays illustrent une hétérogénéité du prix
de l’éducation et donc de l’investissement en capital humain. La structure proposée par ces
10
Introduction générale
auteurs présente néanmoins des limites puisque elle ne prend pas en compte les différences
d’intensités factorielles entre les deux secteurs.
Nous pouvons souligner que cette désagrégation de la production en plusieurs secteurs
est également fondée lorsque l’on s’intéresse aux émissions de pollution. En effet, ces
facteurs génèrent des effets externes spécifiques. Le capital physique est identifié comme
un facteur de production plus polluant que le capital humain. Les secteurs du bâtiment et
des transports, intensif en capital physique, sont ceux qui émettent le plus de pollution.
0.2 L’éducation comme enjeu politique
La tendance à l’accroissement des dépenses en éducation ces vingt dernières années
confirme une volonté des décideurs politiques de soutenir l’intervention des gouverne-
ments dans la sphère éducative. Les organismes mondiaux prônent d’ailleurs une inter-
vention massive des états dans l’éducation primaire, particulièrement pour les pays en
développement.
De nombreux arguments sont évoqués pour le soutien à l’éducation, les plus récur-
rents étant l’effet positif sur la croissance économique, l’amélioration de la distribution
des revenus, et la correction des inefficiences de marché.
0.2.1 Intervention publique en éducation
Synthétiquement, les dépenses éducatives peuvent être soutenues par trois institutions :
le marché, lorsque l’agent contracte un crédit pour financer ses études, la famille, ou bien
l’Etat. L’accès au marché du crédit pour financer l’investissement privé en capital humain
existe mais reste marginal et il concerne principalement les derniers stages d’apprentis-
sage. La famille, comme nous l’avons vu, joue un rôle déterminant favorisant ainsi une
11
Introduction générale
reproduction des inégalités. Enfin, le rôle de l’Etat dans la sphère éducative est multiple et
répond souvent à des objectifs d’équité ou d’efficience. Il peut fournir une éducation pu-
blique, c’est le cas de l’éducation primaire et secondaire dans la majorité des pays, ou bien
subventionner des dépenses éducatives privées. Ce deuxième schéma correspond alors
plus à un soutien au financement de l’éducation supérieure. Une vaste littérature, large-
ment documentée par Glomm et Ravikumar dans les années 90, s’est focalisée sur le rôle
et l’effet des politiques publiques en éducation, sans pour autant prendre en considération
les spécificités sectorielles énoncées dans la section précédente.
De nombreuses formalisations de l’intervention gouvernementale dans le financement
de l’éducation sont proposées dans la littérature. Partant du constat que l’éducation est un
outil de politique économique largement utilisé, une première approche formalise une po-
litique éducative exogène et examine ses impacts sur la croissance ou les inégalités. Bräu-
ninger and Vidal (2000) par exemple, considèrent une subvention aux dépenses d’éduca-
tion privée pour étudier le lien entre politique éducative et croissance tandis que Glomm
and Ravikumar (1997) ou Blankenau and Simpson (2004) supposent que le gouverne-
ment alloue une part fixe de son PIB au financement de l’éducation publique. Par ailleurs,
d’autres économistes considèrent une éducation publique qui résulte de choix privés. Les
agents déterminent, à travers un vote, le montant de la taxe qui sera implémentée par le
gouvernement pour financer l’éducation. Dans ce cas, dès lors que les agents sont suppo-
sés homogènes, un système d’éducation exclusivement privé ou publique est équivalent :
les agents allouent le même montant aux dépenses éducatives. Dans un cadre avec agents
hétérogènes, ces deux systèmes diffèrent. De nombreux auteurs se sont ainsi penchés sur
la question du financement de l’éducation en confrontant éducation publique et privée
(Glomm and Ravikumar, 1992, Zhang, 1996, Cardak, 2004). 3
3. Dans Glomm and Ravikumar (1992) et Zhang (1996), un système privé puis public est étudié suc-
12
Introduction générale
D’après De Fraja (2002), considérer que la politique éducative est déterminée de façon
endogène par un vote des agents n’est pas réaliste. Les choix politiques d’un agent ne sont
pas principalement motivés par le programme de politique éducative proposé par le gou-
vernement. De Fraja (2002) adopte un point de vue plus normatif, en considérant que la
politique éducative est un outil dont le montant doit être déterminé de façon endogène par
le décideur politique. Il propose ainsi un schéma de politique éducative permettant de dé-
centraliser la solution d’un planificateur social utilitariste. Il identifie trois raisons qui jus-
tifient l’intervention publique en éducation : la distribution des revenus, les imperfections
de marchés (l’accès à l’emprunt pour financer l’éducation peut être contraint), et enfin les
externalités générées par l’accroissement du niveau de qualifications. Comme mentionné
précédemment, de nombreux effets externes associés à l’éducation sont identifiés. Dans
la théorie économique, l’existence d’externalités positives en éducation est un argument
central pour justifier l’action publique dans la sphère éducative. 4 En effet, les agents n’éva-
luent pas les bénéfices sociaux de leur investissement, un soutien public à l’éducation est
donc requis pour lever la sous-optimalité de l’équilibre de marché. Des travaux de Cremer
et al. (2005), Cremer and Pestieu (2006) ou Docquier et.al (2007) ont adopté une approche
similaire. Cremer and Pestieu (2006) soulignent que lorsque l’éducation est financée par
des parents altruistes, une subvention n’est pas toujours positive car elle résulte de deux
forces opposées. D’un côté les agents, dotés d’un altruisme de type “joy of giving”, valo-
risent uniquement leurs enfants et non pas l’ensemble des générations futures. D’un autre
côté, les motivations altruistes de l’agent ne sont pas forcément prises en compte par le
cessivement. Cardak (2004) considère que les deux systèmes coexistent et l’agent peut voter pour l’un oul’autre des financements.
4. Empiriquement, la mesure des externalités, et donc du rendement social de l’éducation, reste com-plexe, malgré le développement continu des méthodes économétriques ainsi que l’accès à de nouvelles basesde données. L’intervention des gouvernements pour soutenir l’éducation reste néanmoins largement souhai-tée et justifie les nombreuses études qui s’efforcent d’identifier les effets des politiques éducatives.
13
Introduction générale
planificateur social. En effet, ils considèrent l’argument de Harsanyi (1995) selon lequel
ce type de motivation ne doit pas être inclus dans une fonction de bien-être car ce sont
des préférences “externes”, qui conduisent à donner un poids différent aux individus en
fonction du nombre de bienfaiteurs qu’ils ont. Les recommandations de politiques faites
par Cremer and Pestieu (2006) peuvent toutefois être remises en question dans la mesure
où ils font abstraction de la croissance endogène et de l’existence d’une fonction d’ac-
cumulation de capital humain. Au contraire, le modèle de croissance endogène proposé
par Docquier et.al (2007) permet de conclure qu’une subvention à l’éducation positive est
toujours requise.
À la lecture de ces travaux une question émerge : Le gouvernement doit-il encoura-
ger l’investissement éducatif en subventionnant les dépenses privées d’éducation ou bien
fournir une éducation publique pour atteindre ses objectifs ? Les réponses fournies dans la
littérature dépendent du cadre d’étude choisit. L’implémentation d’une éducation publique
soulève des problèmes d’effets d’éviction lorsque celle-ci est combinée avec des dépenses
éducatives privées. L’impact de l’éducation publique va ainsi fortement dépendre de la
substituabilité entre des deux dépenses. Dans un modèle où éducation publique et privée
sont parfaitement substituables, la décentralisation de la solution du first-best ne requiert
pas d’éducation publique. Les études de Cremer et al. (2005) et Cremer and Pestieu (2006)
montrent que celle-ci peut être souhaitable lorsqu’une solution de second rang est consi-
dérée. En l’absence d’altruisme et lorsque les agents ne disposent pas des outils financiers
pour investir dans l’éducation, l’éducation publique est évidemment requise (Boldrin and
Montes, 2005).
Bien que l’influence positive de l’éducation publique sur l’accumulation de capital
humain est largement admise, certains auteurs soulignent que la prise en compte de son
14
Introduction générale
financement, afin d’évaluer pleinement les effets d’une politique éducative, peut aboutir
à des conclusions différentes. Blankenau and Simpson (2004) considèrent différentes mé-
thodes de financement afin d’explorer la relation croissance-éducation publique. Ils isolent
l’effet positif direct de la politique et les ajustements de l’équilibre général. L’effet direct
dépend uniquement de la capacité des dépenses publiques à créer du capital humain tandis
que les ajustements de l’équilibre général vont varier en fonction du niveau des dépenses,
de la méthode de financement, et des paramètres technologiques dans la fonction d’accu-
mulation du capital humain. Ils soulignent ainsi le caractère non-monotone de la relation
entre croissance et éducation publique pour la plupart des schémas de financement retenus.
Nous montrerons dans le chapitre 2 de la thèse que cette relation dépend également des
préférences des agents, dès lors qu’un financement par des taxes sectorielles est utilisé.
0.2.2 Education et perspective internationale
Face à l’importance grandissante des unions économiques, formées par des pays dont
les caractéristiques en termes de choix et de politiques éducatives diffèrent, il nous semble
fondé de considérer les questions relatives à l’éducation dans un contexte international. Au
sein de l’Union Européenne (UE), les décisions de politique éducative se font au niveau
national, il incombe ainsi à chaque Etat de définir sa politique en matière d’éducation.
Néanmoins, l’UE a défini une politique de base en matière d’éducation, l’article 149 du
traité de Lisbonne stipule que “la communauté contribue au développement d’une édu-
cation de qualité en encourageant la coopération entre Etats membres et, si nécessaire,
en appuyant et en complétant leur action [...]”. Ainsi, différents dispositifs d’échanges et
de mobilité entre pays ont été mis en place afin de favoriser les externalités éducatives
transfrontalières. Cela nous amène à penser que la prise en compte d’une dimension inter-
15
Introduction générale
nationale pour évaluer le rôle de l’intervention publique dans l’éducation est importante.
Au préalable, il est évidemment essentiel de déterminer comment le processus d’accu-
mulation de connaissances d’un pays évolue dans un contexte international. La littérature
fournit d’intéressantes contributions à ce sujet.
Les implications de l’intégration économique ont d’abord été étudiées dans un cadre
où le capital physique était le seul facteur de production accumulé de façon endogène.
Nous définissons ici l’intégration économique comme l’intégration du marché du capital
physique, de sorte que l’économie intégrée forme un seul marché commun caractérisé par
une mobilité parfaite des capitaux. Suite à l’intégration, il y a un flux de capitaux du pays
ayant un faible taux d’intérêt vers le pays ayant un taux d’intérêt élevé, jusqu’a ce que
le rendement du capital s’égalise entre les deux économies. Galor (1992b) a examiné les
conséquences de cette intégration entre deux pays dont les préférences pour le temps, et
donc les taux d’épargne, diffèrent. Il montre que l’intégration est bénéfique pour le pays
impatient et dommageable pour l’autre. L’analyse est différente lorsque l’on considère le
capital humain. La mobilité internationale des travailleurs reste marginale, même au sein
des unions économiques. 5 Par conséquent, les études portant sur l’intégration économique
dans un cadre avec capital humain supposent généralement que ce facteur est imparfaite-
ment mobile, voir immobile entre pays (voir par exemple De la Croix and Monfort, 2000,
Michel and Vidal, 2000, Viaene and Zilcha, 2002, Egger et al., 2010). De plus, face à la
volonté des gouvernements de favoriser les échanges et la mobilité des étudiants, il est ad-
mis que des effets externes spécifiques au capital humain peuvent être considérés. Lorsque
deux économies sont intégrées, les dépenses éducatives d’un pays affectent l’accumulation
de capital humain de l’autre pays. C’est cette approche que privilégient Michel and Vidal
5. Selon une étude de la commission européenne de 2006 sur la mobilité professionnelle, moins de 2%de citoyens européens vivent dans un pays différent de leur pays d’origine.
16
Introduction générale
(2000) pour examiner comment l’intégration affecte la croissance de long-terme des éco-
nomies. Dans un modèle de croissance endogène, ils étendent l’analyse de Galor (1992b),
en soulignant le rôle joint des différences de patience et d’altruisme entre les pays, qui
conduisent respectivement l’accumulation de capital physique et humain. Comme dans le
modèle de Galor (1992b), la mobilité parfaite du capital physique va engendrer une éga-
lisation des taux d’intérêt entre les pays. Celle-ci va s’accompagner d’une convergence
des taux de croissance dès lors qu’il existe des externalités inter-régionales en éducation.
Les préférences des agents en éducation et en épargne de chaque pays vont conditionner
les bénéfices et les coûts de l’intégration. Au regard de ce résultat, nous pouvons conclure
que, dans un perspective de long-terme, des mesures visant à favoriser la création d’exter-
nalités entre régions peuvent s’avérer être un outil de politique économique efficace pour
favoriser la convergence de pays membres d’une union économique.
Par ailleurs, des auteurs ont porté une attention particulière aux conséquences de l’inté-
gration internationale du marché des capitaux sur l’investissement en éducation des agents
et des gouvernements. Viaene and Zilcha (2002) montrent que l’intégration modifie la po-
litique publique “optimale” en éducation, visant à maximiser le bien-être des travailleurs.
Plus récemment, Egger et al. (2010) valident empiriquement que les entrées de capitaux
favorisent l’investissement en éducation supérieure.
Certains économistes ont adopté une perspective quelque peu différente en considérant
l’effet du commerce international sur la formation du capital humain. Galor and Mount-
ford (2008) ont fait valoir l’idée que la distribution du capital humain au niveau mondial
résultait des effets du commerce international. La structure du commerce détermine les
investissements en éducation dans les pays de sorte que l’accumulation de capital humain
tend à être stimulée dans les pays développés et à être freinée dans les pays en développe-
17
Introduction générale
ment.
L’ensemble de ces études suggère que le développement des échanges internationaux
altère les incitations à s’éduquer ainsi que les politiques éducatives mises en place par les
gouvernements au niveau local. Néanmoins, les contributions dans ce domaine ne nous
paraissent pas suffisantes dans la mesure où les spécificités sectorielles, évoquées précé-
demment, ne sont pas considérées. De plus, ces travaux négligent la nature non échan-
geable des dépenses éducatives. La chapitre trois de la thèse s’appliquera à proposer un
cadre d’analyse qui prend en compte ces caractéristiques.
0.3 L’éducation un enjeu pour l’environnement
“Education is one of the most powerful tools for providing individuals with the appro-
priate skills and competencies to become sustainable consumers. UNESCO has designated
2005-2014 as the Decade of Education for Sustainable Development, to which the OECD
will contribute by highlighting good practices in school curricula for sustainable develop-
ment.” OECD (2008), p.26.
0.3.1 Intervention publique et gestion des problèmes environnementaux
Depuis les années 1970, période de l’adoption du premier programme d’action pour
l’environnement au niveau de l’UE, l’ensemble des gouvernements doit composer avec
la gestion des problèmes environnementaux pour atteindre leurs objectifs de croissance.
La politique environnementale et ses enjeux sont désormais au centre des débats natio-
naux et internationaux. La nécessité d’une croissance économique durable, favorisant une
réduction des émissions de pollution et une baisse de l’utilisation des énergies non renou-
velables, est clairement admise.
18
Introduction générale
En plus de la mise en place de règlementations visant à réduire les émissions pol-
luantes, une approche économique de la politique environnementale consistant à utiliser
des “instruments économiques”, s’offre aux décideurs politiques pour parvenir à atteindre
les objectifs environnementaux. Cette approche économique a fortement progressé ces der-
nières années dans la plupart des pays développés et en développement. 6 Le recours aux
instruments économiques est en partie motivé par leur caractère incitatif, qui tend à mo-
difier durablement les comportements des producteurs et des consommateurs. Dans cette
perspective, Schubert (2009) mentionne que la politique économique est essentielle pour
fournir les incitations à la réduction de l’utilisation des énergies fossiles.
Le décideur politique dispose de plusieurs outils économiques pour diminuer les émis-
sions de pollution. Le plus communément utilisé est la taxe sur les activités polluantes.
Celle-ci revêt plusieurs formes et offre l’avantage de fournir des recettes fiscales qui
peuvent être réinjectées dans l’économie. De nombreuses fonctions ont été attribuées à
ces recettes fiscales. Parmi les plus récurrentes, il est souvent mentionné qu’elles peuvent
être utilisées pour raffermir le solde budgétaire, pour augmenter certaines dépenses (dans
le domaine de l’environnement ou non), ou encore pour réduire d’autres taxes. L’argument
d’un double dividende, selon lequel la taxe génère un bénéfice environnemental et écono-
mique 7 est ainsi fréquemment évoqué (voir Chiroleu-Assouline, 2001, Schob, 2005, pour
des revues de la littérature sur l’hypothèse du double dividende). Plus récemment, Schu-
bert (2009) suggère que les recettes de la taxe doivent être redistribuées de façon forfaitaire
aux ménages, dès lors que les firmes répercutent le coût de la taxe sur leur prix.
6. La base de données collectée par l’OCDE sur les instruments de politiques environnementales utilisésdans les pays de l’OCDE et dans vingt pays hors-OCDE (http ://www2.oecd.org/ecoinst/queries/) illustreaisément ce phénomène.
7. Le bénéfice économique peut prendre plusieurs formes : une baisse du chômage, un effet distributifou encore une baisse de la distorsion fiscale. Ce dernier point est souvent étudié et repose sur l’idée d’untransfert de la pression fiscale du travail vers les activités polluantes.
19
Introduction générale
Les objectifs environnementaux sont désormais couplés aux objectifs de croissance des
économies. Cela a conduit les économistes théoriques à examiner les effets des politiques
environnementales dans des modèles de croissance endogène. Parmi les contributions les
plus influentes, nous pouvons citer les travaux de Bovenberg and Smulders (1996) ou Ono
(2003a) et Ono (2003b), qui s’intéressent aux coûts et aux bénéfices, en terme de crois-
sance, d’une taxe sur la pollution. L’originalité de l’étude de Bovenberg and Smulders
(1996) réside dans la prise en compte d’un progrès technologique endogène. Selon eux, la
qualité de l’environnement agit comme un facteur de production public qui vient accroitre
la productivité des travailleurs. Ainsi, dans un modèle où le flux de pollution est un input de
la fonction de production, l’amélioration de la productivité peut permettre de surpasser les
effets négatifs directs de la politique environnementale. Les travaux d’Ono fournissent une
contribution intéressante à cette littérature, tout en soulignant l’intérêt des modèles à géné-
rations imbriquées pour aborder les questions environnementales. Ils mettent en évidence
un effet positif supplémentaire de la taxe environnementale, provenant de son caractère
intergénérationnel. En diminuant le flux de pollution, une taxe verte améliore la qualité
de l’environnement léguée aux générations futures. L’amélioration de l’environnement
va agir comme un effet revenu positif sur la génération suivante. Celle-ci va ainsi pou-
voir allouer plus de ressources à l’investissement en capital, au profit de la croissance de
long-terme. En utilisant des modèles de croissance endogène avec capital humain, d’autres
auteurs ont conclu qu’un accroissement de la taxe sur la pollution pouvait favoriser l’édu-
cation au détriment des activités polluantes. Dans Grimaud and Tournemaine (2007), ce
résultat émane de l’accroissement du prix relatif du bien polluant suite à sa taxation. Dans
Pautrel (2011), l’hypothèse d’un secteur d’abattement plus intensif en capital humain que
le secteur du bien final est requise pour observer cette relation.
20
Introduction générale
Dans ces études, les auteurs ne se préoccupent pas du recyclage du produit de la taxe,
excluant ainsi de potentiels bénéfices économiques. Les études prenant en compte l’utili-
sation des recettes fiscales se concentrant principalement sur le cas où un accroissement
de la taxe sur la pollution est compensé par une baisse de la taxe sur le travail et sont peu
développée dans des modèles de croissance endogène. Nous souhaitons élargir ces ana-
lyses en proposant un recyclage du revenu plus en accord avec les récentes informations
rapportées par l’OCDE, tout en considérant le caractère endogène de la croissance.
Comme nous l’avons souligné, les gouvernements disposent d’une large catégorie
d’instruments politiques pour parvenir aux objectifs environnementaux et économiques,
et la combinaison de différents instruments est aujourd’hui largement promut par les or-
ganismes internationaux. D’après le rapport de l’OECD (2007), les deux arguments prin-
cipaux avancés en faveur de policy mix sont : (1) les problèmes environnementaux ont de
multiples aspects et externalités, qui requièrent des politiques différentes, et (2) les ins-
truments politiques peuvent mutuellement se renforcer. De plus, le rapport de l’OECD
(2010), stipule que l’utilisation d’une taxe sur la pollution n’engendrera pas forcement les
effets escomptés si elle n’est pas complétée par d’autres politiques, visant notamment à
sensibiliser les agents à l’impact de leurs comportements sur l’environnement. Nous ver-
rons que l’éducation peut être un instrument efficace pour atteindre cet objectif.
Dans ce contexte, il nous parait ainsi important de proposer un cadre d’analyse per-
mettant d’évaluer les conséquences macroéconomiques d’un policy mix visant à favoriser
la croissance économique et l’environnement.
21
Introduction générale
0.3.2 La prise en compte d’une conscience environnementale
La commission européenne a récemment mis en exergue les préoccupations grandis-
santes des ménages concernant les problèmes environnementaux et le rôle qu’ils peuvent
jouer dans la protection de l’environnement. 8 Les agents seraient ainsi dotés d’une conscience
environnementale individuelle, qui capture leur capacité à adopter des comportements éco-
logiques.
Dans la théorie économique, Forster (1973) est le premier à proposer une formalisation
qui prend en compte les préoccupations environnementale des agents, en considérant une
économie où la pollution est dommageable pour le bien-être. Le consommateur maximise
une fonction d’utilité qui dépend de sa consommation et de la qualité de l’environnement,
et dévoue ainsi une part de son revenu à des activités de maintenance qui vont permettre
de diminuer la pollution. Cette idée a largement été reprise dans la littérature et le papier
célèbre de John and Pecchenino (1994) a permis de mettre en lumière l’incidence de ces
comportements verts sur les générations futures. Dans cet article, ils considèrent un cadre
à générations imbriquées, où l’environnement est un bien public multidimensionnel qui
fournit un indice général de la qualité de l’environnement. Ils formalisent la sensibilité
aux préoccupations environnementales en supposant que les agents altruistes valorisent
la qualité de l’environnement qu’ils vont laisser à leurs enfants. L’analyse de John and
Pecchenino (1994) suggère que les fondements de la courbe de Kuznet environnementale,
qui décrit une relation en U entre développement économique et qualité de l’environne-
ment, réside dans la sensibilité écologique des agents. Durant les premières phases de
développement, l’agent tend à privilégier sa consommation, au détriment des activités de
maintenance. Lorsque la perte d’utilité engendrée par la dégradation de l’environnement
8. Voir, European Commission (2008), “The Attitudes of European Citizens towards the Environment”,Eurobarometer 296, DG Environment
22
Introduction générale
est trop grande, il commence à investir en maintenance. Le développement de l’économie
s’accompagne alors d’une amélioration de l’état de l’environnement.
Dans la majorité des modèles théoriques, la contribution privée de l’agent à la protec-
tion environnementale est motivée par un altruisme paternaliste. Néanmoins, une impor-
tante littérature a révélé que ce type de modèle, dans lequel le donateur se soucie unique-
ment du bien public, ne retranscrit pas correctement le comportement des agents. Andreoni
(1990) propose la formalisation d’un altruisme impur qui puisse expliquer les dons de cha-
rité. Plus précisément, il suppose que l’agent est doté d’un altruisme paternaliste mais qu’il
valorise également sa propre contribution au bien public. La présence d’un altruisme de
type “joy of giving” est pertinente empiriquement lorsque l’on s’intéresse aux comporte-
ments associés à la protection de l’environnement (Menges et al., 2005). Pour modéliser
l’action écologique des agents nous empruntons, par conséquent, la modélisation proposée
par Andreoni (1990) dans le chapitre quatre de la thèse.
0.3.3 Quel rôle pour la politique éducative ?
Face à la mise en évidence du rôle central des consommateurs dans la gestion des
problèmes environnementaux, il apparait évident que les variables pouvant influencer la
conscience écologique des agents représentent un outil de politique puissant à la disposi-
tion des décideurs politiques. Dans cette perspective, l’éducation a été identifiée comme
un facteur déterminant de la conscience environnementale (Blomquist and Whitehead,
1998, Witzke and Urfei, 2001). L’explication de cette relation trouve ses fondements dans
l’argument de Nelson and Phelps (1966), selon lequel l’éducation améliore les capaci-
tés à recevoir, à décoder, et à comprendre l’information. Les agents plus éduqués tendent
ainsi à avoir des préférences vertes plus importantes. L’éducation est désormais considé-
23
Introduction générale
rée par l’ensemble des organismes internationaux comme un moyen de relever les défis
du développement durable. L’assemblée générale des Nations Unis a consacré la décennie
2005-2014 des Nations Unis à l’éducation au service du développement durable. Ces faits
nous amène à nous questionner sur le rôle de la politique éducative dans la gestion des
problèmes environnementaux, dans un contexte où l’éducation détermine les préférences
vertes des agents. Nous trouvons une illustration pertinente de cette question dans la lit-
térature. Prieur and Bréchet (2013) examinent les effets d’une politique éducative lorsque
les préférences environnementales d’un agent dépendent de son niveau d’éducation. En
considérant que l’agent a une utilité pour sa consommation et pour la qualité de l’envi-
ronnement future, ils définissent la conscience environnementale comme l’élasticité entre
ces deux éléments. Les choix éducatifs sont ignorés, les dépenses éducatives résultent en
effet d’une politique publique exogène. Leur résultat central remet en cause l’idée qu’une
croissance du capital humain, favorable à la croissance économique, est compatible avec
un accroissement de la qualité environnementale. Dès lors que l’éducation accroit les pré-
férences vertes, un accroissement du capital humain favorise les activités de maintenance
au détriment de l’épargne privée. L’accumulation de capital étant réduite, la convergence
vers un sentier de croissance équilibrée n’est pas garantie, l’économie peut alors converger
vers un état stationnaire, où l’opportunité d’accroitre la croissance économique et l’envi-
ronnement est perdue. Ils soulignent que le gouvernement peut définir une politique édu-
cative permettant d’éviter cette situation. Cette analyse offre un point de vue intéressant
concernant les implications de la conscience environnementale endogène. Nous souhaitons
toutefois l’étendre, en proposant notamment un schéma politique plus riche, afin d’évaluer
les conséquences de l’implémentation d’un policy mix faisant intervenir un politique édu-
cative.
24
Introduction générale
0.4 Présentation générale de la thèse
Cette thèse propose une analyse purement théorique des effets de l’éducation en consi-
dérant des modèles de croissance endogène avec capital humain. L’aspect intergénération-
nel associé à l’éducation nous conduit à privilégier l’utilisation de modèles à générations
imbriquées pour l’ensemble de nos études.
0.4.1 Enjeux de la thèse
Comme nous l’avons vu, la littérature théorique visant à étudier les mécanismes de
formation du capital humain ainsi que les effets des politiques éducatives sur la croissance
est abondante. Conjointement, face à la nécessité de favoriser une croissance économique
durable, les problématiques associant croissance et environnement se sont développées.
Des contributions nous semblent toutefois requises. En effet, certaines spécificités re-
latives à la formalisation de la structure productive de l’économie ne sont pas considérées.
De plus, la prise en compte de l’éducation dans le schéma de politique environnementale
et la considération de préférences vertes endogène restent récentes et peu documentées.
Partant de ce fait, nous souhaitons enrichir la littérature sur la croissance et l’éducation
en proposant des cadres d’analyse plus riches qui nous permettent de traiter de probléma-
tiques non explorées.
Comment les caractéristiques sectorielles d’un pays influencent son schéma de poli-
tique optimale ? Est-ce que des taxes sectorielles peuvent être plus efficaces que des taxes
sur la production agrégée pour financer l’éducation ? Quelles sont les incidences de l’in-
tégration économique sur l’accumulation de capital humain lorsque l’on prend en compte
le caractère non-échangeable de l’éducation ? Comment une politique environnementale
25
Introduction générale
associée à un soutien à l’éducation affecte la croissance lorsque les préférences vertes des
agents sont endogènes ? Le travail proposé dans les chapitres de cette thèse tente de fournir
des éléments de réponses à ces questions.
0.4.2 Résumé des chapitres
Les quatre chapitres proposés dans cette thèse reposent sur des modèles de croissance
endogène avec capital humain. L’accumulation de capital humain est une fonction des
dépenses privées en éducation qui émanent de parents altruistes. En plus de ces caractéris-
tiques communes, des propriétés spécifiques sont considérées dans chacun des chapitres
afin de répondre au mieux à la problématique posée.
Le premier chapitre de cette thèse se focalise sur l’investissement optimal en capital
physique et en capital humain défini par un planificateur social. Due à la présence d’exter-
nalité en capital humain, le rendement social de l’investissement éducatif est supérieur au
rendement privé. De plus, l’investissement en capital physique est inefficace compte tenue
de la durée de vie finie de l’agent. L’intervention du gouvernement est donc souhaitée pour
corriger les défaillances du marché.
Comme nous l’avons précédemment mentionné, il n’existe pas de consensus clair
concernant le traitement des sentiments altruistes dans la fonction de bien-être social (voir
Cremer and Pestieu, 2006). Par conséquent, nous comparons, au préalable, la solution op-
timale définie pas un planificateur utilitariste à celle définie par un planificateur qui ignore
les sentiments altruistes. Nous montrons que le taux de croissance optimal est toujours plus
grand avec une fonction utilitariste mais les différences quantitatives restent faibles. Pour
la suite, nous privilégions l’approche utilitariste car les arguments proposés par Harsanyi
26
Introduction générale
(1995), en faveur de la seconde approche, sont discutables dans notre cadre d’analyse.
Tout d’abord, il propose de n’inclure que les préférences personnelles de l’agent. Dans la
mesure où les enfants ne sont pas en mesure de prendre de décision concernant leur édu-
cation et que l’action altruiste émane des parents, considérer l’altruisme paternaliste pour
l’éducation comme une préférence personnelle nous semble justifié. Ensuite, il stipule que
la prise en compte des sentiments altruistes conduit à donner plus de poids aux agents qui
ont de nombreux bienfaiteurs. Dans notre modèle, chaque agent à un seul donateur, son
parent, ce qui invalide cet argument.
La deuxième partie du chapitre s’applique à définir l’investissement optimal en capital
physique et en capital humain, afin de déterminer conjointement la politique éducative et
les transferts intergénérationnels permettant d’atteindre un équilibre optimal. Nous traitons
cette question dans un cadre à deux secteurs, dans lequel les secteurs d’investissement et
de consommation se différencient en termes d’intensité en capital physique et humain. En
étudiant comment ce différentiel influence la politique optimale, nous étendons l’analyse
uni-sectorielle de Docquier et.al (2007), qui se focalise sur le rôle joué par le taux d’ac-
tualisation social et les préférences pour le temps de l’agent. Dans la mesure où le gouver-
nement doit influer sur l’investissement en capital physique et humain, nous donnons une
attention particulière à l’importance relative de ces deux stocks de capital. Lorsque le ratio
capital physique sur capital humain optimal est plus faible que celui du laissez faire, le
décideur politique doit favoriser l’investissement en capital humain par rapport à l’inves-
tissement en capital physique. Nous montrons que cette situation peut être renversée suite
à une modification du différentiel d’intensité factorielle entre les secteurs. L’étude d’une
telle modification est notamment motivée par les travaux de Zuelta and Young (2012) qui
mettent en évidence que l’apparente stabilité dans la part global des facteurs, observée aux
27
Introduction générale
Etats-Unis, masque des évolutions contrastées des parts de facteurs au niveau sectoriel. En
étudiant précisément l’effet des caractéristiques sectorielles sur la solution optimale, nous
mettons en évidence le résultat suivant : si le secteur d’investissement devient plus intensif
en capital humain, tandis que l’inverse se produit dans le secteur de consommation, alors la
politique optimale devra soutenir plus fortement l’éducation, dès lors que le planificateur
social accorde suffisamment d’importance aux générations futures. Ce résultat s’explique
par le fait que, si les générations futures sont fortement valorisées, le planificateur favorise
le bien d’investissement. Ainsi, les changements dans ce secteur prévalent sur les change-
ments apparaissant dans le secteur de consommation.
Le premier chapitre met en exergue qu’une structure à deux secteurs permet de prendre
en compte des mécanismes ignorés par un modèle uni-sectoriel. Cela nous amène à utiliser
ce cadre d’analyse pour traiter un point central de la littérature sur éducation et croissance,
relatif au financement des dépenses publiques en éducation. L’apport de l’éducation pu-
blique sur l’accumulation de capital humain est aujourd’hui clairement admis. Néanmoins,
lorsque les coûts relatifs à son financement sont pris en compte, des résultats plus contras-
tés émergent concernant l’effet des dépenses publiques en éducation sur la croissance.
L’impact positif direct de la politique peut en effet être réduit si d’autres facteurs influen-
çant la croissance sont affectés négativement. Blankenau and Simpson (2004) montre ainsi
que l’effet d’une augmentation des dépenses publiques en éducation sur la croissance dé-
pend fortement de la structure de financement imposée par le gouvernement, du montant
des dépenses, et des paramètres technologiques dans la fonction d’accumulation de ca-
pital humain. Ils soulignent que si la pression fiscale est fortement dommageable pour
l’épargne, l’impact global d’une politique éducative peut s’avérer négatif. Plus récem-
ment, Basu and Bhattarai (2012) montrent que l’élasticité du capital humain aux dépenses
28
Introduction générale
publiques est le facteur décisif de la relation croissance-dépenses publiques en éducation.
Nous étendons ces études en explorant cette relation dans un modèle à deux secteurs. La
désagrégation de la production entre un secteur de service et un secteur manufacturier nous
permet notamment d’évaluer les implications, en termes de croissance, d’un financement
de l’éducation publique par des taxes sectorielles. Afin de prendre en compte les effets
de l’intervention publique sur les choix privés, nous considérons que l’agent alloue une
part de son revenu à l’épargne et à l’investissement en éducation. La loi d’accumulation
du capital humain dépend donc de l’éducation publique, de l’éducation privée, et du ca-
pital humain de la génération précédente. Le gouvernement alloue une part fixe du PIB à
l’éducation publique et finance cette politique en prélevant, successivement, une taxe sur
la production manufacturière, sur la production de service ou sur la production agrégée.
Nous montrons que les dépenses publiques en éducation favorisent toujours direc-
tement l’accumulation de capital humain mais peuvent réduire les dépenses privées en
épargne et en éducation. Dans un cadre à deux secteurs, l’ampleur de ces deux effets op-
posés dépend fortement des préférences des agents pour le temps, pour l’éducation et pour
les services. Dans un premier temps, nous occultons les différences d’intensité en fac-
teur entre les secteurs afin d’identifier clairement les conséquences d’un financement par
une taxation sectorielle. Comme les biens servant à l’éducation et l’épargne ne sont pas
produits dans le même secteur, une taxe sectorielle modifie le prix relatif de ces deux dé-
penses. Une taxe sur le secteur manufacturier favorise l’accumulation de capital humain
au détriment de l’accumulation de capital physique, tandis qu’une taxe sur le secteur des
services agit dans le sens inverse. Cet ajustement du prix relatif réduit ou renforce les bé-
néfices de l’éducation publique, en fonction des préférences des agents. En considérant le
cas où l’éducation est financée par une taxe sur le secteur manufacturé, nous montrons le
29
Introduction générale
résultat suivant : pour atteindre un taux de croissance maximal, une économie qui se carac-
térise par de faibles préférences pour les services devra allouer une part plus grande de son
PIB à l’éducation publique qu’une économie où les agents consomment principalement
des biens manufacturés. De plus, nous montrons que l’utilisation d’une taxe sectorielle
plutôt que d’une taxe sur l’ensemble de la production peut permettre d’atteindre un taux
de croissance plus élevé. Par exemple, une taxe sur le secteur des services sera plus per-
formante si les préférences pour les services sont assez importantes. De cette façon, la
taxe est suffisamment faible et son impact négatif est moindre que celui provenant d’une
taxe agrégée. Dans un deuxième temps, nous considérons que le secteur des services est
intensif en capital humain. Le prix relatif des biens est alors déterminé par l’équilibre sur
le marché des biens non-échangeables. Dans ce cas, les préférences des agents influencent
la relation entre éducation publique et croissance même si la politique est financée par une
taxe agrégée.
Dans le troisième chapitre de cette thèse, nous abordons une perspective différente
en considérant le cadre d’une économie intégrée. Nous souhaitons examiner l’impact de
l’intégration économique pour des pays dont les choix en termes éducatifs diffèrent. Les
modèles à deux secteurs utilisés dans les chapitres précédents sont particulièrement adap-
tés pour la formalisation d’un contexte international, notamment par ce qu’ils vont nous
permettre de prendre en compte l’existence de biens non-échangeables. Nous étendons
ainsi les études sur cette question, notamment l’étude de Michel and Vidal (2000) qui
se limite à un cadre uni-sectoriel. En accord avec les faits, nous considérons désormais
que le bien produit dans le secteur des services a un caractère non-échangeable. La pré-
sence d’un bien non-échangeable introduit une variable clef dans notre analyse : le prix
relatif des biens non-échangeables en termes de biens échangeables, qui va déterminer le
30
Introduction générale
rendement des facteurs de production. En supposant que le secteur non-échangeable est
intensif en capital humain, le taux de croissance va dépendre positivement de ce prix rela-
tif. En outre, la décomposition de la production en deux secteurs nous permet de prendre
en compte la productivité globale des facteurs (PGF) de chaque secteur. Cette distinction
est pertinente car de récentes études ont montré que la faible PGF observée dans les pays
en développement par rapport aux pays développés était principalement due à leur faible
productivité dans le secteur échangeable (Hseih and Klenow, 2008).
L’intégration économique résulte d’une mobilité parfaite des capitaux physique entre
les pays. De plus, nous supposons qu’il existe des externalités transfrontalières en capital
humain. Dans le contexte du processus de Bologne, la mobilité des étudiants au sein de
l’Union Européenne a fortement progressé durant la dernière décennie. Cela nous conduit
donc à supposer que lorsque deux économies sont intégrées, les dépenses éducatives d’un
pays affectent l’accumulation de capital humain dans l’autre pays.
Notre étude introduit une dimension supplémentaire par rapport aux travaux de Michel
and Vidal (2000). En considérant une économie domestique et une économie étrangère
qui s’intègrent, la dynamique de transition dans l’économie intégrée n’est pas uniquement
conduite par la dynamique du ratio entre dépenses éducatives du pays domestique et du
pays étranger. Elle dépend également de celle du prix relatif du bien non échangeable en
termes de bien échangeable. 9 L’intégration génère un ajustement transitoire du prix relatif
dans les deux pays, qui se répercute sur le taux de croissance des économies. Cet ajuste-
ment dépend principalement des différences de PGF entre les secteurs échangeables. En
effet, ces différences vont conditionner le différentiel de taux d’intérêt entre les pays et
donc les mouvements de capitaux lors de l’intégration. En supposant que le secteur non-
9. L’égalisation des taux d’intérêts dans l’économie intégrée se traduit pas un rapport des prix relatifsentre les pays qui est constant. La dynamique du prix relatif est donc identique dans les deux pays.
31
Introduction générale
échangeable est intensif en capital humain, l’intégration économique génère une chute du
prix relatif et donc du taux de croissance dans le pays ayant une faible PGF dans le secteur
échangeable. Cet effet est néanmoins transitoire, à long terme les taux de croissance des
deux économies convergent.
Concernant les effets de long-terme de l’intégration, nous nuançons les conclusions
qui émanent d’une analyse uni-sectorielle. Dans un cadre à un secteur, le résultat suivant
émerge : si l’économie étrangère est caractérisée par une préférence pour l’éducation et un
degré de patience plus faible que l’économie domestique, alors l’intégration sera toujours
bénéfique pour le pays étranger et coûteuse pour le pays domestique. Dans un cadre à deux
secteurs, les bénéfices de l’intégration dépendent également des préférences pour le bien
non-échangeable et des différences de PGF sectorielles, de sorte que le résultat précédent
ne tient pas toujours.
L’étude menée dans ce chapitre nous permet de formuler des recommandations poli-
tiques. Tout d’abord, nous soulignons qu’il est important d’évaluer les coûts et bénéfices
de l’intégration économique à court et à long terme, dans la mesure où les conclusions
peuvent diverger selon l’horizon considéré. Ensuite, les décideurs politiques doivent pour-
suivre leurs efforts pour favoriser les externalités en capital humain entre pays car elles
permettent d’assurer une convergence des économies à long terme et d’éviter une dyna-
mique non-monotone du taux de croissance.
Face à la contrainte grandissante que représente l’évolution de la pollution sur la crois-
sance, nous décidons de considérer dans le quatrième et dernier chapitre de cette thèse, un
modèle de croissance endogène avec environnement. Nous modélisons l’évolution de l’en-
vironnement selon John and Pecchenino (1994). L’objectif de ce chapitre est d’examiner
si un policy mix, faisant intervenir une politique de soutien à l’éducation et à l’environ-
32
Introduction générale
nement, peut favoriser une croissance durable. À la différence des chapitres précédents,
nous considérons une structure productive à un secteur, afin de nous concentrer sur l’évo-
lution des comportements des consommateurs. En effet, nous apportons une attention par-
ticulière aux préférences vertes des agents, qui reflètent l’importance qu’ils accordent à
l’environnement. Ces dernières évoluent au cours du temps et de récentes études nous en-
seignent que l’éducation tend à favoriser les comportements écologiques. Nous supposons
que l’agent, en plus de valoriser l’éducation de ses enfants, est doté d’un altruisme impur
(Andreoni, 1990) pour l’environnement, qui motive ses dépenses en maintenance privée.
Le poids de cet altruisme dans la fonction d’utilité de l’agent fournit une mesure de la
conscience environnementale. Nous supposons ainsi que ce paramètre est influencé posi-
tivement par le niveau d’éducation de l’agent et par le stock de pollution dans l’économie.
Nous proposons un schéma de politiques économiques en accord avec les récentes recom-
mandations faites par l’OCDE et évaluons ses incidences sur le court et le long-terme. De
façon standard, le décideur politique lève une taxe sur la production polluante. Le revenu
de cette taxe est réinjecté dans l’économie et va alimenter deux dépenses. Une dépense en
éducation, visant à encourager l’investissement privé en capital humain, et une dépense
en maintenance publique, qui va venir améliorer l’indice de qualité environnementale. La
subvention à l’éducation va avoir un impact positif sur l’environnement, dans la mesure
où un accroissement de l’éducation favorise le revenu et donc les dépenses en activités de
maintenance. Le rôle positif de l’éducation dans la sphère environnementale ne se limite
néanmoins pas à ce mécanisme, puisque la sensibilité aux problèmes environnementaux
va être stimulée par le soutien à l’éducation. La part allouée à chacun des postes de dé-
penses est supposée constante, de sorte que le gouvernement dispose de deux instruments
de politiques.
33
Introduction générale
L’analyse dynamique nous permet de prouver qu’un sentier de croissance équilibrée
durable, où le capital humain et l’indice de qualité environnementale croissent au même
taux, existe sous certaines conditions. Ces conditions nécessitent notamment que les acti-
vités de maintenance soient suffisamment efficaces. La convergence vers cet état de long-
terme peut se faire de façon cyclique en raison des préférences vertes endogènes. En effet,
l’évolution de la conscience environnementale au cours du temps tend à affecter le trade-
off entre dépenses en éducation et en maintenance. La théorie nous enseigne que de tels
cycles illustrent des inégalités intergénérationnelles. Nous montrons qu’un accroissement
de la taxe environnementale peut alimenter ces cycles si le gouvernement n’alloue pas cor-
rectement le revenu de la taxe. De la même façon, à long terme, la taxe permet de converger
vers un taux de croissance durable plus élevé seulement si le gouvernement recycle cor-
rectement ses recettes fiscales. Nous pouvons finalement conclure qu’il existe un policy
mix approprié qui garantit un meilleur taux de croissance sans favoriser l’émergence de
cycles.
34
Chapter 1
Optimal human and physical capital accumulation
Part A Social optimum in an OLG model with paternalistic
altruism
1.1 Introduction
How does the inclusion of paternalistic altruistic feelings in the social planner’s ob-
jective affect the optimal growth rate ? According to a strict definition, and assuming that
utility is cardinal and unit-comparable (but not level-comparable) between generations,
the social utility is the discounted sum of individual utility functions, namely a Utilitarian
social function. However, a study of Harsanyi (1995) on theory of morality recommends
to exclude all external preferences, as altruism, from the social utility function (Harsa-
nyi social function). 1 Indeed, based on Dworkin (1977, p.234), Harsanyi differences bet-
ween “personal” and “external” preferences. The first refer to preferences for enjoyment
of goods, while the second to assignment of goods and opportunities of others. From his
1. Harsanyi (1995) proposes a theory of morality in order to determine what we should do to have a goodlife from a moral point of view ? He defines the best moral value are those “likely to produce the greatestbenefits for a society as a whole” (p.324) and argues that “the arithmetic mean of all individuals utilityfunction in this society” (p.324) is a way to measure the welfare of a society.
36
1.1 Introduction
point of view, the altruistic feelings enter in the second class of preferences, which should
be excluded from the social utility function. He mentions two arguments for this : First,
we should treat agents in accordance with their own personal preferences, i.e in the way
they want to be treated rather than in accordance with the way other people want them to
be treated. Second, when external preferences are included, the weight assigned to each
individual differ according to the number of well-wishers and friends that agent have.
Thus there is no consensus yet on the correct way to write this social utility function.
On one side, considering an overlapping generations one sector model with consumption
separable utility function, Cremer and Pestieu (2006) underline that optimal policy de-
pends on the specification of the social utility function. Nevertheless, they do not clearly
examine the implications on the optimal balanced growth path. On the other side, De la
Croix and Monfort (2000) do not include the “joy of giving” term in the welfare function.
In this paper, we show that the way to write the central planner objective is crucial for the
optimal growth path. In this purpose, we consider the example of a paternalistic altruism
where agents are concerned by the level of human capital of their children. We also assume
that human capital is a simple function of parents investment in their child’s education.
The aim of this paper is to explore the consequence to consider one or the other of the
social welfare function, when we determine the benchmark optimal solution that will be
compared to the laissez-faire. Our contribution is to quantitatively investigate the conse-
quences of omitting the altruistic term in the social utility function. As long as education
is ignored in the social utility function, the only way to increase welfare is to maximize
consumption. Conversely, when child’s education provides direct welfare to parents, there
is an arbitrage in the social utility function between consumption and education. This is the
reason why the relationship between human capital and capital intensity depends on pre-
37
1 Optimal human and physical capital accumulation
ference parameters with the Utilitarian social function. We show that the optimal growth
rate is higher with the Utilitarian social function than in the Harsanyi social function. We
calibrate the model to quantify the difference between Utilitarian and Harsanyi optimal
paths.
1.2 Social optimum and paternalistic altruism
Consider a perfectly competitive economy in which the final output is produced using
physical capital K and human capital H . The production function of a representative
firm is an homogeneous function of degree one : F (K,H). We assume for simplicity a
complete depreciation of the capital stock within one period. Denoting, for any H 6= 0,
k ≡ K/H the physical to human capital ratio, we define the production function in inten-
sive form as f(k).
Assumption 1. f(k) is defined over R+, Cr over R++ for r large enough, increasing
(f ′(k) > 0) and concave (f ′′(k) < 0). Moreover, limk→0 f′(k) = +∞ and limk→+∞ f ′(k) =
0.
We can also compute the share of capital in total income :
s(k) = kf ′(k)f(k)
∈ (0, 1) (1.1)
As in Michel and Vidal (2000), we consider a three-period overlapping generations model.
In their first period of life, individuals are reared by their parents. In the second period, they
work and receive a wage, consume, save and rear their own children. In their last period of
life, they are retired and consume their saving returns. Following Glomm and Ravikumar
(1992), we consider a paternalistic altruism according to which parents value the quality
38
1.2 Social optimum and paternalistic altruism
of education received by their children. Thus, the preferences of an altruistic agent born in
t− 1 are represented by :
Ut = u(ct, dt+1) + v(ht+1) (1.2)
where ct and dt+1 represent adult and old aggregate consumption, and ht+1 child’s human
capital.
For the discussion of the balanced growth paths, we formulate an assumption similar
to that of Docquier et.al (2007) and Del Rey and Lopez Garcia (2012) :
Assumption 2.
i) u(c, d) is C2, increasing with respect to each argument (uc(c, d) > 0, ud(c, d) >
0), concave and homogeneous of degree a, with a ∈ ]0, 1[. Moreover, for all d > 0,
limc→0 uc(c, d) = +∞, and for all c > 0, limd→0 ud(c, d) = +∞.
ii) v(h) is C2, increasing (v′(h) > 0), concave and homogeneous of degree a, with
a ∈ ]0, 1[. Moreover limh→0 v′(h) = +∞. 2
To guarantee the existence of a balanced growth path in growth model, the utility function
has to be homothetic. Moreover, it has to be strictly concave to ensure that the planner’s
solution is well behaved. Thus, to satisfy these properties, we assume a utility function
homogeneous of degree a with 0 < a < 1.
Parents devotes et unit of good to his child’s education, so human capital in t + 1 is
given by :
ht+1 = G(et) (1.3)
2. Note that Assumption 2 ii) implies that v(h) is a power function : y = xa.
39
1 Optimal human and physical capital accumulation
Assumption 3. The human capital production function G(e) is strictly increasing and
linear with e. 3
As et is in units of final good, this assumption combined with the homotheticity of the
utility function and the linear homogeneity of the production function implies that the
optimal solution for et is linearly homogeneous in ht. This guarantees the presence of
inter-temporal spillovers in human capital. Moreover, the model generates endogenous
growth since the human capital production function has constant returns to scale.
At the decentralized equilibrium grandparents’ expenditures in education generate a
positive intergenerational external effect in human capital accumulation. Indeed, parents
decide for their child’s education but do not consider the impact of this decision on their
grand child’s education. We assume that population is constant over time and is normalized
to 1, i.e. Nt = N = 1. 4 Moreover, clearing condition on the labor market gives Ht =
Nht = ht and thus Kt = htkt.
The social planner maximizes the discounted sum of the life-cycle utilities of all cur-
rent and future generations under the resource constraint of the economy and the human
capital accumulation equation.
maxct,dt,Kt+1,Ht+1
∞∑
t=−1
δt (u(ct, dt+1) + ǫv(ht+1)) (1.4)
3. We choose this simple linear function for the sake of simplicity but our results stay valid if we assumea human capital accumulation function concave in et :
ht+1 = hµt e
1−µt 0 < µ < 1
4. The assumption of stationary population is innocuous for our analysis. It just implies that along abalanced growth path per capita variables, c, d, and h grow at the same rate than the two stocks of capital Kand H .
40
1.2 Social optimum and paternalistic altruism
subject to F (Kt, Ht) = ct − dt − et −Kt+1
ht+1 = G(et)(1.5)
with δ ∈ (0, 1), and ǫ taking alternatively the extreme values 0 (Harsanyi social function)
and 1 (Utilitarian social function).
The Lagrange function is :
L = δ−1[u(c−1, d0) + ǫv(h0)]
+∞∑
t=0
δt(
u(ct, dt+1) + ǫv(ht+1) + qt(
htf(kt)− ct − dt −G−1(ht+1)− ht+1kt+1
))
(1.6)
where qt the Lagrange multiplier associated with the constraint. First order conditions forall t > 0 are :
uc(ct, dt+1) = qt (1.7)
ud(ct−1, dt) = δqt (1.8)
δt+1f ′(kt+1)qt+1 = δtqt (1.9)
δtǫv′(ht+1) + δt+1qt+1(f(kt+1)− kt+1f′(kt+1)) = δtqtG
−1′(ht+1) (1.10)
htf(kt)− ct − dt −G−1(ht+1)− ht+1kt+1 = 0 (1.11)
41
1 Optimal human and physical capital accumulation
limt→∞
δtqtKt+1 = 0, limt→∞
δtqtHt+1 = 0 (1.12)
where equation (1.10) is obtained by differentiating the Lagrangean with respect to ht+1
and making a simplifying substitution using equation (1.9). For initial conditions c−1, K0
and H0 and for all t > 0, optimal solutions satisfy equations (1.7) to (1.12).
From (1.7), (1.8) and (1.9) we can rewrite the condition that gives optimal physical capital
accumulation :
f ′(kt+1) =uc(ct, dt+1)
ud(ct, dt+1)=
qtδqt+1
(1.13)
From (1.7), (1.9) and (1.10), we obtain the following expression that determines optimal
human capital accumulation :
MRSe/c = G−1′(ht+1)− kt+1
(
1
s(kt+1)− 1
)
(1.14)
with MRSe/c ≡ ǫv′(ht+1)uc(ct,dt+1)
the marginal rate of substitution between education and first
period consumption.
Equation (1.14) displays the main difference between the two approaches. With Har-
sanyi social function (ǫ = 0), equation (1.14) becomes :
f ′(kHt+1) = (f(kHt+1)− kHt+1f′(kHt+1))G
′ (eHt) (1.15)
with kH and eH respectively the capital intensity and education spending in the Harsanyi
case. The optimal return on investment in human capital (through education) is equal to the
return on physical capital since the central planner does not differentiate between physical
42
1.3 Optimal growth and capital accumulation
and human capital accumulation. The welfare increases only with consumption. Thus,
along the balanced growth path, defined by a constant physical to human capital ratio,
the optimal kH corresponds to the standard Modified Golden Rule. Conversely, with the
Utilitarian social function (ǫ = 1), from equation (1.14) we get :
f ′(kUt+1) > (f(kUt+1)− kUt+1f′(kUt+1))G
′ (eUt) (1.16)
with kU and eU respectively the capital intensity and education spending in the Utilitarian
case. The optimal return on investment in human capital (through education) is lower than
the one on physical capital because human capital accumulation provides direct welfare.
There is a trade off between consumption and education. Then, we depart from the Modi-
fied Golden Rule through the MRS term. In this latter case, preferences (time preference,
altruism) affect the relationship between human capital and capital intensity, whereas they
do not in the Harsanyi social utility function.
1.3 Optimal growth and capital accumulation
The previous section has shown that qualitatively the way we specify the social utility
function matters for capital accumulation. We determine here precisely optimal path for
capital accumulation kt and ht. We know from k0 given and equations (1.7) and (1.9), that
optimal physical capital intensity K/H is constant from t = 1. Thus, the social planner
has to determine the initial stocks H1 and K1 which drive the economy along the optimal
path. Along this optimal path, K and H will grow at a constant rate g.
Proposition 1. If G (h) = bh, 1 ≥ b > 0, the optimal path is determined by g∗ =
43
1 Optimal human and physical capital accumulation
[δf ′(k1)]1
1−a − 1 with k1, q0 and h1 solutions of
h1k1 = h0f (k0)− c0(k1, q0)− d0(q0)−h1b
(1.17)
ǫv′ (h1) + q0k1
(
1
s (k1)− 1
)
=q0b
(1.18)
f ′(k1)
(
1
b+ k1
)
− f(k1) = ǫ
(
Ψ(k1)
1 + Ψ(k1)
)1−a
Θ(k1) (1.19)
with c−1, h0, k0 predetermined, d0(q0) solution of q0 = ud (c−1, d0) /δ and
c0(k1, q0) ≡ φ−1 (f ′ (k1))
[
uc(φ−1(f ′(k1)),1)q0
] 11−a
with φ (x) = uc (x, 1) /ud (x, 1)5,
Ψ(k1) = (1 + g∗(k1))φ−1((1 + g∗(k1))
1−a/δ) and Θ(k1) =[f(k1)−(k1+ 1
b )(1+g∗(k1))]1−a
v′(1)
δuc
(
1, 1φ−1(f ′(k1))
) .
Proof. As marginal utility of consumption is homogeneous of degree (a− 1), from (1.7)
and (1.9), we have
δf ′(k1) = (1 + g∗)1−a (1.20)
Equations (1.17) and (1.18) in Proposition 1 comes from equations (1.10) and (1.11)
at time t = 0. From homogeneity (Assumption 1) and equations (1.7) at time t = 0
and (1.8) at time t = 1, we get c0 = φ−1(
q0δq1
)
d1. Substituting in (1.7) gives
da−11 uc
[
φ−1(
q0δq1
)
, 1]
= q0. As from equation (1.9) at time t = 0 gives f ′ (k1) =
q0δq1
, we finally get the result for c0(k1, q0). Equation (1.19) is obtained using the
first order conditions along the balanced growth path. Using homogeneity of u,
equations (1.7) and (1.8) and balanced growth path properties, according to which
ct−1
dt−1= ct
dtand dt−1
dt= dt
dt+1, we have c1 = (1 + g∗(k1))φ
−1((1 + g∗(k1))1−a/δ)d1.
5. Under assumption 2, function φ(.) is invertible.
44
1.3 Optimal growth and capital accumulation
Adding equation (1.11) and homogeneity of v, we can define the following relation-
ship, c1h1
=[f(k1)−(k1+ 1
b )(1+g∗(k1))](1+g∗(k1))φ−1((1+g∗(k1))1−a/δ)
1+(1+g∗(k1))φ−1((1+g∗(k1))1−a/δ). Finally, using equation
(1.10), we obtain the last equation of the system, from which we get k1.
Proposition 1 describes the general system whose resolution gives the optimal growth
rate and capital accumulations. We wish to show that the results obtained in terms of op-
timal growth and stocks highly depends on the way the social utility function is specified.
Indeed, with the Harsanyi function, the relationship between human capital and capital in-
tensity do not depend on time preference and altruism whereas they do with the Utilitarian
social function.
For simplicity let us consider the following assumption :
Assumption 4. Utility is characterized by u(c, d) = ca + βda and v(h) = γha, 0 < a <
1, 0 < γ < 1 and technologies are given by f(k) = kα.
Proposition 2. Under assumption 4, there exists a unique value k1i, i = U,H, satisfying
equation (1.19). Moreover, k1H > k1U , hence optimal growth rate is always higher in the
Utilitarian case.
Proof. Under assumption 4, equation (1.19) becomes
Ω1(k1) = Ω2(k1) (1.21)
with :
Ω1(k1) ≡αkα−1
1i
b− (1− α)kα1i (1.22)
and
45
1 Optimal human and physical capital accumulation
Ω2(k1) ≡γǫ
δ
kα1i −(
δαkα−11i
) 11−a (k1i + 1/b)
1 + (β/δ)1
1−a
1−a
=αkα−1
1i
b−(1−α)kα1i (1.23)
We have limk1→0
Ω1(k1) = +∞, limk1→+∞
Ω1(k1) = −∞, dΩ1(k1)/dk1 < 0, and Ω1(k1) =
0 for a unique value k1 = k1 with k1 ≡ α(1−α)b
. Concerning Ω2(k2), limk1→0
Ω2(k1) =
−∞, limk1→+∞
Ω2(k1) = +∞ and for Ω2(k1) > 0, dΩ2(k1)/dk1 > 0. Thus Ω2(k1) = 0
for a unique value k1 = k1. Moreover the sign of Ω2(k1) is given by the sign of the
term[
1−(
δαaα(b(1− α))a(1−α))
11−a
]
, which is always positive, and then k < k.
We deduce finally that, when ǫ = 0, Ω2(k1) = 0, and the unique solution to equation
(1.21) is k1H = k1, and when ǫ = 1, the unique solution to equation (1.21) is k1U ∈
[k1, k1H ].
The function Ω2(k2) is strictly increasing in ǫ while Ω1(k1) does not depend on this
parameter. Considering the properties of these two functions given in Proposition 2, this
implies that there is a negative relationship between k1 and ǫ. The higher the weight that
the social planner gives to altruistic feelings, the higher the growth rate. 6
Quantitatively, the spread between Utilitarian and Harsanyi optimal paths may be large.
We show this through a numerical example calibrated on five countries using proxies for
time preference (β) based on Wang et al. (2011), for altruism (γ) based on Armellini and
Basu (2010) (using data from European and World Value Survey four-wave data-file 1981-
2004, 2006). For the social planner discount rate (δ), we use the average real interest rate
for 1980-2010 (World Development Indicators, World Bank 2010) following Armellini
and Basu (2010). The proxy for δ in country i is the ratio between the country i average
6. We can note that, even if the growth rate does not directly depend on the technology parameter b, thisparameter affects negatively the variable k1, in the Utilitarian and the Harsanyi cases. As the growth rate isdecreasing with k1, the technology parameter b affects positively the growth rate.
46
1.3 Optimal growth and capital accumulation
real interest and the Russian real interest rate (which is the highest) multiplied by 0.95
which is lower than one to guarantee the convergence of the social objective.
Calibrations. Consider the specific example given in Assumption 4 to calibrate our mo-
del. Table 1.1 collects the parameter values. We are interested in highlighting the
spread between the optimal paths obtained with the Harsanyi and Utilitarian social
functions. Table 1.2 compares optimal paths in the two cases
Country Time Preference (β) Altruism Degree (γ) Social discount Rate (δ)Germany 0.8 0.587 0.91
Japan 0.87 0.435 0.37Russian Federation 0.77 0.563 0.95
United States 0.84 0.758 0.64
TABLE 1.1 – Parameters Values
Country g∗U/g∗H
Germany 1.023Japan 1.228
Russian Federation 1.017United States 1.151
TABLE 1.2 – Calibration results
Table 1.2 shows that there are important differences between the two specifications
of the social utility function. From Proposition 2, we know that the optimal growth rate
is always higher with the Utilitarian social function. The way we specify the social uti-
lity function matters for the determination of the optimal growth paths. For example, in
the United States, the Utilitarian approach leads to an optimal growth which is 15.15%
higher than the one emerging with the Harsanyi utility function. We can remark that the
ratio g∗U/g∗H differs a lot between countries. This is due to cross-country heterogeneities
in agent’s preferences and social discount factor, that shape the optimal growth rate in the
47
1 Optimal human and physical capital accumulation
Utilitarian case. The discount factor explains the main differences between countries. The
intuition is the following : the higher the weight that the social planner gives to future ge-
nerations, the lower the influence of private preferences on the optimal solution. Therefore,
when δ is high the Utilitarian solution is very close to that of the Harsanyi function.
When altruistic components are not included in the social welfare function, the opti-
mal solution is affected, so does the comparison between the equilibrium and the optimal
allocation. Consider an optimal policy that consists to correct the market inefficiencies of
the laissez-faire economy. The way we write the social welfare function does not change
the nature of inefficiencies. In our overlapping generation setting, an education policy
is required to correct intergenerational externalities in education while intergenerational
transfers have to be implemented to correct inefficiency in physical capital accumulation.
Our analysis highlights that an optimal education policy would reach a human capital with
the Utilitarian approach which is higher than the one obtained with the Harsanyi utility
function.
48
Part B Should a country invest more in human or physical ca-
pital ? A two-sector endogenous growth approach
1.4 Introduction
Should a country invest more in human or physical capital ? This paper addresses this
issue considering a two-sector two-factor overlapping generations growth model.
The optimal amount a country should invest in human or physical capital is usually
analyzed through a basic aggregated one-sector framework, even though economists agree
that this approach is too restrictive to describe the production process. Representing the
whole economy through a one-sector structure does not allow sectoral differences and
relative price adjustments between sectors to be considered. Empirical evidence suggests
that sectoral relative prices vary (see Hseih and Klenow, 2008), and differences occur
especially between rich and poor economies. Herrendorf and Valentinyi (2008) also show
that factor intensity is sector-dependent in the US economy. Zuelta and Young (2012)
emphasize that the US labor income share within the agricultural and manufacturing sector
fell between 1958 and 1996 whereas it increased in the service sector. This means that the
apparent stability of the US global labor share hides contrasted evolutions of sectoral labor
shares. In a recent study, Takahashi et.al (2012) measure the capital intensity difference
between consumption and investment good sectors in the post-war Japanese economy and
49
1 Optimal human and physical capital accumulation
in other main OECD countries. Before the 1973 oil shock, the Japanese investment sector
was capital intensive. They observe a capital intensity reversal after the oil shock, with the
consumption sector becoming capital intensive compared to the investment sector. They
suggest that a capital intensive investment sector in Japan before 1973 may explain the
high speed growth observed over this period, as suggested by the Rybczynski theorem 1.
A rise in physical capital endowment increases production in the investment goods sector
more than in the consumption goods sector. As the investment sector produces physical
capital, this leads to a magnification effect. 2 According to Takahashi et.al (2012), a one-
sector framework fails to account for this phenomenon.
The aim of this paper is to show that, in a two-sector framework with endogenous
growth driven by human capital 3, public policy recommendations depend on the diffe-
rential of factor intensities between sectors. We depart from the Glomm and Ravikumar
(1992) overlapping generations (OLG) model introducing a two-sector two-factor produc-
tion structure à la Galor (1992a). Most papers using a two-sector model of endogenous
growth with education consider a sector producing a good which can either be consumed
or invested, and an education sector (Bond et al., 1996, Mino et.al, 2008). 4
We distinguish between a consumption and an “investment” sector which use both
human and physical capital. In this specific setting, the good produced in the investment
sector is used for education spending and investment. Due to the two-sector structure, the
1. The theorem states that a rise in the endowment of one factor will lead to a more than proportionalexpansion of the output in the sector which uses that factor intensively, at constant relative goods price.
2. Unlike Japan, in other OECD countries, the consumption sector is capital intensive. There is neither amagnification effect nor capital intensity reversal.
3. Galor and Moav (2004) show that, in the process of development, growth is first driven by physicalcapital accumulation and then by human capital accumulation. We focus on the stage of growth when humancapital matters.
4. Bond et al. (1996) use a continuous-time model and Mino et.al (2008) use a discrete-time modelwith infinitely-lived agents. To our knowledge, in the literature, the two-sector two-factor formalization witheducation sector and final good sector is not used in the OLG model with education spending.
50
1.4 Introduction
relative price between the two goods plays a crucial role in the factor allocation between
sectors. It matters both for human capital accumulation and for economic growth. We ana-
lyze optimal physical and human capital accumulation and explain how these allocations
depend on sectoral characteristics.
This paper characterizes the socially optimal balanced growth path in a two-sector
framework with paternalistic altruism. The social optimum is defined by a social planner
who maximizes the discounted sum of utility of all future generations. We prove that it
crucially depends on the sectoral differences in terms of capital intensities. Consider two
laissez-faire (LF) economies with the same characteristics, except for relative factor inten-
sities. These economies may generate physical capital excess or scarcity, with respect to
the optimum, even if the global factor share is identical between these two countries. In
a one-sector model, the sectoral capital intensities differential would be ignored : optimal
factor accumulation would be the same for these two economies.
We define a Relative Factor Accumulation (RFA) reversal as a situation where a change
in sectoral capital intensities makes the optimal global capital intensity higher (lower) than
the LF when it was initially lower (higher) than the LF. For example, in a country where
the optimal global physical to human capital ratio is lower than the LF, an optimal po-
licy would favor human capital investment. If sectoral changes lead to an RFA reversal -
which means that the optimal global physical to human capital ratio becomes higher than
the competitive one - the optimal policy would be to favor physical capital accumulation.
We emphasize that such RFA reversal may arise depending on the level of individuals’
impatience. Then, to achieve the first-best, the government should consider a relationship
between these sectoral capital intensities and time preference. To sum up, relative capital
intensity between sectors is crucial to determine the scheme of optimal policy. Consider
51
1 Optimal human and physical capital accumulation
that investment sector becomes more human capital intensive while the opposite is ob-
served in the consumption sector : as long as future generations are sufficiently valued,
authorities have to adapt their policy in favor of education.
The remainder of this paper is organized as follows. In Section 1.5, we set up the
theoretical model. Section 1.6 is devoted to the planner’s solution. In Section 1.7, we
compare optimal and laissez-faire solutions. In Section 1.8, we examine the design of
optimal policy. Finally, Section 1.9 concludes.
1.5 The Model
1.5.1 The production structure
We consider an economy producing a consumption good Y0 and a capital good Y1. 5
Each good is produced using physical capital Ki and human capital Hi, with i = 0, 1,
through a Cobb-Douglas production function. The representative firm in each industry
faces the following technology :
Y0t = A0Kα00t H
1−α00t (1.24)
Y1t = A1Kα11t H
1−α11t (1.25)
α1, α0 ∈ (0, 1) A1 > 0, A0 > 0
Full employment of factors holds so that, K0t+K1t = Kt, and H0t+H1t = Ht, where Kt
and Ht are respectively the total stock of physical capital and the aggregate human capital
at time t.
5. The two-sector formalization that we consider refer to a non-durable (consumption) and a durable(capital) good.
52
1.5 The Model
We denote the physical to human capital ratio in sector i by ki = Ki/Hi, and the share
of human capital allocated to sector i, hi = Hi/H and obtain :
y0t = A0kα00t ; y1t = A1k
α11t (1.26)
h0k0 + h1k1 = k, h0 + h1 = 1 (1.27)
where y0t and y1t are the production per unit of human capital, Yi/Hi, in each sector. First
order conditions of the firm’s problem give :
wt = (1− α1)A1kα11t = P0t(1− α0)A0k
α00t (1.28)
Rt = α1A1kα1−11t = P0tα0A0k
α0−10t (1.29)
where wt represents the wage, Rt the rental rate of capital, and P0 the relative price of the
consumption good in terms of the investment good. From (1.28) and (1.29), we derive the
physical to human capital ratios as functions of the price of the consumption good :
k0t = B(
α0(1−α1)α1(1−α0)
)
(P0t)1
α1−α0
k1t = B(P0t)1
α1−α0
with B =(
α0
α1
)
α0α1−α0
(
A0
A1
) 1α1−α0
(
1−α1
1−α0
)
α0−1α1−α0
(1.30)
In this model, there are as many mobile factors as sectors, so that factor returns depend
only on the relative price and do not depend on the global capital intensity k.
For simplicity, we assume a complete depreciation of the capital stock within one period,
such that the stock of physical capital in t+ 1 is equal to investment in physical capital in
t.
53
1 Optimal human and physical capital accumulation
1.5.2 Household’s behavior
The economy is populated by finitely-lived agents. In each period t, N persons are
born, and they live for three periods. Following Glomm and Ravikumar (1992), we consi-
der a paternalistic altruism whereby parents value the quality of education received by
their children. In their first period of life, agents get educated. In their second period of
life, when adult, they are endowed with ht+1 efficiency units of labor that they supply in-
elastically to firms. Their income is allocated between current consumption, saving and
investment in children’s education. 6 As we assume no population growth, we normalize
the size of a generation toN = 1. In their third period of life, when old, agents retire. They
consume the proceeds of their savings.
The preferences of a representative agent born at time t − 1 are represented by a log-
linear utility function :
U(ct, dt+1, ht+1) = (1− β) ln ct + β ln dt+1 + γ lnht+1 (1.31)
0 < β < 1 ; 0 < γ < 1
where β/(1 − β) is the private discount factor 7, γ the degree of altruism, ct and dt+1
correspond respectively to adult and old aggregate consumption, and ht+1 child’s human
capital. Parents devote et to their children’s education. Human capital in t+1 is given by :
6. We do not focus on the trade-off between public and private education funding, but only assess theimpact of sectoral factor intensity differences on the design of optimal policy. Consequently, we assumeeducation is only private. Since agents are homogeneous, a private funding system is equivalent to a purelypublic regime financed by a proportional income tax chosen by parents.
7. The discount factor β/(1− β) is equal to 1/(1 + ρ) with ρ the rate of time preference.
54
1.5 The Model
ht+1 = bet (1.32)
Notice that even if there is no explicit externality in our simple human capital production
function, grandparents’ expenditure in education generate a positive intergenerational ex-
ternal effect. 8 Indeed, parents decide for their child’s education but do not consider the
impact of this decision on their grandchild’s human capital. The more educated children
are, the more they earn when adults and invest in their own children’s education.
An agent born at date t − 1 maximizes his utility function over his life cycle, with
respect to budget constraints and human capital accumulation function :
maxst,et
U(ct, dt+1, ht+1) = (1− β) ln ct + β ln dt+1 + γ lnht+1
s.t wtht = P0tct + st + et (a)
stRt+1 = P0t+1dt+1 (b)
ht+1 = bet (c)
First order conditions give the optimal education et, and the optimal saving st :
et =γ
1 + γwtht (1.33)
8. We consider that human capital accumulation function is linear in educational expenditure, as in theautarky model of Michel and Vidal (2000). This formalization allows us to keep analysis tractable, especiallyfor the planner’s solution.
55
1 Optimal human and physical capital accumulation
st =β
1 + γwtht (1.34)
1.5.3 Equilibrium
Since the size of the working age population is equal to one, the effective labor supply
at time t is Ht = ht. The clearing condition in the capital market is Kt+1 = st, and as by
defintion Kt+1 = kt+1Ht+1, we have :
kt+1 =stht+1
(1.35)
Using (1.32), (1.33), (1.34) and (1.35), we obtain the equilibrium physical to human
capital ratio which is constant over time.
kt+1 =β
bγ= kLF (1.36)
The consumption market clearing condition in period t is :
ct + dt = Y0t
ct + dt = A0k0tα0h0tht
(1.37)
We define the growth rate as the growth rate of human capital :
1 + gt =ht+1
ht(1.38)
To highlight the impact of sectoral differences, we define ε, the factor intensity dif-
56
1.5 The Model
ferential between consumption and investment sectors : ε ≡ α0 − α1. Therefore, when
ε tends to zero, we have two identical sectors. The larger ε, the larger the differences
between sectors. Hereafter, we focus on the Balanced Growth Path (BGP).
Definition 1. We define a Balanced Growth Path as an equilibrium where all variables
grow at a constant and same rate 9, such that the per-unit-of-effective-labor variables are
constant.
We can now characterize the economy’s growth rate :
Lemma 1. On the balanced growth path, the growth rate is
1 + g =γ
(1 + γ)bA1w (1.39)
with
w = (1− α1)kα11
and
k1 =β
bγ(1− α1)
α1(1 + γ)(1− α1 − ε)
ε(1− β) + α1(1 + γ)(1.40)
Proof. See Appendix 1.10.1.
9. The linear homogeneity of the production function implies that the growth rate along the BGP isidentical for all per human capital variables
57
1 Optimal human and physical capital accumulation
The equilibrium growth rate depends on the degree of altruism γ and on the private
discount factor β, in the same way as in Michel and Vidal (2000). 10 When altruism tends
to zero, agent does not value the human capital of his child and he stops to invest in human
capital. Thus, the growth rate tends to zero as well because private education spending is
the only variable input in the production of human capital. In a two-sector framework, the
growth rate is also shaped by the spread between sectoral factor intensities ε, through the
wage.
Proposition 3. For α1 given, the competitive growth rate is decreasing with ε.
Indeed, for α1 given, sectoral capital intensity k1 is a decreasing function of ε. A rise in
ε (meaning a rise in α0, when α1 is given) leads to a decrease in the marginal productivity
of human capital in the consumption sector. Due to factor mobility between sectors, mar-
ginal productivity of human capital is decreasing as well in the investment sector 11, and
so does the growth rate. As a result, when the investment sector becomes more intensive
in physical capital relative to the consumption sector, the economic growth rate goes up.
This is consistent with Takahashi et.al (2012), who show that the investment good sector
was capital intensive with respect to the consumption good sector (ε < 0) during the high
speed growth period in Japan.
1.6 The social planner’s problem
The social planner adopts a utilitarian viewpoint and maximizes the discounted sum of
all future generations’ utilities while allocating output between the different activities. 12
10. Growth rate is an increasing function of β and an increasing and then decreasing function of γ.11. Combining (1.28) and (1.29), we have 1−α1
α1k1 = 1−α0
α0k0
12. We decide to consider an utilitarian view rather than the “Harsayni social welfare function” presentedin the Part A of Chapter 1, because his argument are disputable in our context. First, he mentions that we
58
1.6 The social planner’s problem
The maximization is subject to the clearing conditions on both good markets, the human
capital accumulation equation, and the full employment of resources. In this two-sector
setting, the planner has to allocate both capital stocks between the two sectors at the initial
period (t = 0). We thus have two additional constraints with respect to the one sector case,
corresponding to the full employment of resources at time 0. The planner’s program is
then given by :
maxct,dt,K0t,K1t,H0t,H1t,K0t+1,K1t+1,H0t+1,H1t+1
∞∑
t=−1
δt ((1− β) ln ct + β ln dt+1 + γ lnht+1)
(1.41)
subject to: ∀t > 0 A0Kα00t H
1−α00t = ct + dt
A1Kα11t H
1−α11t = et +Kt+1
ht+1 = bet
Kt = K0t +K1t
ht = H0t +H1t
K0 = K00 +K10
h0 = H00 +H10
K0, h0, and c−1 given
We use the method of the infinite Lagrangian to characterize the optimal solution. The
should include only personal preferences in the social welfare function. Nevertheless, since children arenot able to take decision concerning their education at the lower stage, altruistic feelings can be consideredas personal preferences. Second, when altruism is included, Harsayni argues that the weight assigned toeach individual differs according to the number of well-wishers and friends. In our context, parents act fortheir children, thus all agent have the same number of well-wisher. Moreover, in our definition of the socialwelfare, we do not restrict the weight to be equal for all generations, thus this argument is weak in ourcontext.
59
1 Optimal human and physical capital accumulation
Lagrangian expression can be written as :
L = δ−1((1− β) ln c−1 + β ln d0 + γ lnh0) +∞∑
t=0
δt ( (1− β) ln ct + β ln dt+1 + γ lnht+1) +
∞∑
t=0
δt(
q0t(
A0Kα00t H
1−α00t − ct − dt
)
+ q1t
(
A1Kα11t H
1−α11t − H0t+1 +H1t+1
b−K0t+1 −K1t+1
))
(1.42)
where 0 < δ < 1 is the discount factor, reflecting the social planner’s time preference, q0
and q1 are multipliers associated with the resource constraints in both sectors.
We denote by a superscript asterisk (*) the optimal solution. The maximum of L withrespect to ct, dt, K0t, K1t, H0t, H1t, K0t+1, K1t+1, H0t+1, and H1t+1, is reached when thefollowing conditions are fulfilled for t > 0 :
c∗t =(1− β)
q∗0t; d∗t =
β
δq∗0t(1.43)
δq∗0t+1α0A0k∗
0t+1α0−1 = q∗1t (1.44)
δq∗1t+1α1A1k∗
1t+1α1−1 = q∗1t (1.45)
δq∗0t+1(1− α0)A0k∗
0t+1α0 =
q∗1tb
− γ
h∗t+1
(1.46)
δq∗1t+1(1− α1)A1k∗
1t+1α1 =
q∗1tb
− γ
h∗t+1
(1.47)
q∗00α0A0k∗
00α0−1 = q∗10α1A1k
∗
10α1−1 (1.48)
60
1.6 The social planner’s problem
q∗00(1− α0)A0k∗
00α0 = q∗10(1− α1)A1k
∗
10α1 (1.49)
A0kα00t
∗h∗0th∗
t = c∗t + d∗t (1.50)
A1k∗
1tα1h∗1th
∗
t = h∗t+1
(
1
b+ k∗0t+1h
∗
0t+1 + k∗1t+1h∗
1t+1
)
(1.51)
Transversality conditions are :
limt→∞
δtq∗1tK∗
t+1 = 0, limt→∞
δtz∗t h∗
t+1 = 0 (1.52)
where z∗ is the shadow price of human capital. For initial conditions c−1, K0, h0 and for
all t > 0 an optimal solution is defined as satisfying equations (1.43) to (1.52). 13
Eliminating shadow prices from the first order conditions (FOC) and rearranging the
terms, we obtain the conditions that characterize optimal solutions. From the FOC (1.43),
sectoral implicit prices must equal the marginal utility of consumption. The intertemporal
allocation of consumption between the two goods is given by :
βc∗t = δ(1− β)d∗t (1.53)
Optimal allocation is such that, worker’s marginal utility for consumption and retired
person’s marginal utility for consumption - discounted by the factor δ - are the same.
From (1.43), (1.44) and (1.48), we obtain the optimal growth rate of consumption in t +
13. Since the problem is concave, these first order conditions (FOC) are sufficient.
61
1 Optimal human and physical capital accumulation
1 : 14
c∗t+1
c∗t− 1 =
k∗α0−10t+1
k∗α0−10t
δα1A1k∗α1−11t − 1 (1.54)
We analyze the optimal solution along the BGP and derive the following Lemma :
Lemma 2. Along the BGP, welfare is maximized according to the modified golden rule,
δf ′(k∗1) = 1 + g∗. There exists a unique steady state k∗1 :
k∗1 =α1
b(1− α1)
S + δγ(1− α1)
S + γ(1− δα1)(1.55)
with S = ((1 − β)δ + β)(ε(δ − 1) + 1 − α1)) > 0. Moreover, ∂g∗
∂γ> 0, ∂g∗
∂ε> 0 and
∂g∗
∂β< 0.
Proof. See Appendix 1.10.2.
According to Lemma 2, the optimal growth rate corresponds to the modified golden
rule of the investment good sector. Due to the decreasing returns in physical capital ac-
cumulation, there is a negative relationship between the optimal growth rate g∗ and the
sectoral factor intensity k∗115. Lemma 2 is thus based on the relationships between γ, ε , β
and k∗1 .
The negative impact of altruism factor γ over k∗1 is intuitive : the more altruistic indi-
viduals are, the higher the optimal level of human capital and the lower the ratio k∗1 .
14. Integrating equation (1.43), we havec∗t+1
c∗t=
q∗0tq∗0t+1
. With (1.44) and (1.45) we obtain,c∗t+1
c∗t=
q∗1tq∗0t+1
α1A1k∗
1tα1−1
α0A0k∗
0tα0−0 . Then, with (1.44) we deduce (1.54).
15. Along the BGP, growth rate is given by 1 + g∗ = δA1α1k1∗α1−1, hence it is decreasing in k∗1 .
62
1.6 The social planner’s problem
The negative relationship between ε and k∗1 is due to the assumption of perfect factor
mobility between sectors. A rise in ε corresponds to a rise in α0 at α1 given. When ε
increases, the marginal productivity of capital in the consumption good sector increases,
and then k∗1 decreases to maintain the marginal productivity equality between sectors. This
positive relationship between ε and the growth rate g∗ contrasts with what we obtain in the
competitive equilibrium case : a rise in ε decreases the competitive growth rate.
Concerning the impact of patience, an increase in β means that the planner will in-
vest more in physical capital (increasing k∗1), as it becomes relatively more valuable with
respect to human capital.
Using the equilibrium in the investment sector (1.51), along the BGP we have :
h∗1 =(1 + g∗)(1
b+ k∗)
A1k1∗α1
(1.56)
From equations (1.27), (1.44) to (1.47) 16, and (1.56), we can write k∗ as a function of k∗1
and obtain the following global physical to human capital ratio :
k∗ =b(ε+ α1)(1− α1)k
∗1 − δα1ε
α1b((1− α1) + ε(δ − 1))(1.57)
Through equations (1.55), (1.56) and (1.57), optimal capital intensity (k∗), and optimal
factor allocation between sectors (k∗1 and h∗1) depend on agent’s preferences and ε. 17 The
optimal ratio k∗ is an increasing function of ε. As a result, a change in the spread between
sectoral factor intensities affects the optimal allocation of factors between sectors (k∗0 and
k∗1) and the optimal factor accumulation (k∗).
16. Equations (1.44) to (1.47) gives the standard relationshipk∗
1t+1
k∗
0t+1=(
1−α0
1−α1
)
α1
α0.
17. In this model, the presence of altruism in the utility function makes the optimal physical to humancapital ratio dependent on agent’s time preferences (see Davin et.al, 2012).
63
1 Optimal human and physical capital accumulation
1.7 Laissez-faire and the social optimum
We are interested in the trade-off between investment in education and investment in
physical capital. In our framework, the relative physical to human capital investment in
the economy is given by the physical to human capital ratio, k. We focus on the role of
the two-sector feature on the determination of the optimal level k∗ that is : should the
government invest more in physical capital or in human capital ?
We know from the previous section that the optimal ratio k∗ is affected by the factor
intensity differential ε. As mentioned in the Introduction, empirical studies show that an
apparent constant average factor share may hide sectoral factor share (α0 and α1) changes.
Let us consider the average capital share α = (α0 + α1)/2, and the deviation from this
average, Υ, with Υ ∈ [max−α, α − 1 ; minα, 1 − α] 18. The consumption and
investment sector physical capital shares are respectively α0 = α + Υ and α1 = α − Υ,
and the factor intensity differential ε = 2Υ. This section aims to examine how changes
in sectoral factor shares affect the ratio k∗, whereas the average capital share remains the
same. We obtain the optimal physical to human capital ratio k∗ as a function of Υ :
Lemma 3. The optimal physical to human capital ratio is given by :
k∗ =(α +Υ(1− 2δ)) [Υ(ψ(2δ − 1) + δγ) + (1− α)ψ] + δγ [(1− α)α−Υ(1 + α(1− 2δ)]
b [Υ(ψ(2δ − 1) + δγ) + (1− α)ψ + γ(1− δα)] (1− α−Υ(1− 2δ))
(1.58)
18. We consider average capital share rather than the capital share of the aggregate economy (α). UsingDuregon (2004), the aggregate elasticity of substitution between human and physical capital correspondingto our two-sector framework is σ = P0y0H0+y1H1
P0y1y0k(y1k0h0 + y0k1h1P0), which is not necessarily unitary.
Since α = 1 + σkf ′′(k)/f ′(k), with f(k) the aggregate production function per unit of human capital,aggregate capital share is not analytically tractable for our analysis.
64
1.7 Laissez-faire and the social optimum
where ψ = ((1− β)δ + β).
From this Lemma we underline the non-trivial effects of sectoral factor share move-
ments on the aggregate physical to human capital ratio.
Proposition 4. When δ > 1/2, k∗ is decreasing in Υ. When δ < 1/2, there exists a β such
that when β > β (resp. β < β), k∗ is increasing (resp. decreasing) in Υ.
With β = δ(2δ − 1 + γ)/(1− 2δ)(1− δ).
Proof. see Appendix 1.10.3.
Let us consider δ > 1/2, meaning that the planner favors future generations, and thus
the production of the investment good. An increase in Υ corresponds to a decrease in α1,
the investment good sector becomes human capital intensive and the planner invests more
in education. This is also true when altruism γ is high, whatever δ. Finally, the planner
invests less in physical capital if agents are impatient (β low).
Proposition 4 shows that the optimal strategy of investment depends not only on the
spread of sectoral capital intensities but also on preferences. The optimal response to a
change in factor intensity (Υ) is to invest more in education (resp. physical capital) when
δ and/or γ are high (resp. low) and β is low (resp. high) enough.
Due to externalities, the physical to human capital ratio in the decentralized economy
(kLF ), given by (1.36), differs from the first-best. The positive externality in education
entails an under-accumulation of human capital in the laissez-faire, whereas we can ob-
serve either an under- or an over-accumulation of physical capital. As a result, the com-
petitive ratio k may be higher or lower than the first-best k∗. A higher (lower) optimal
capital intensity than the competitive one means that there is an under-accumulation (over-
accumulation) of physical capital. Optimal policy should favor investment in infrastructure
65
1 Optimal human and physical capital accumulation
(education) rather than in education (infrastructure). Regarding the optimal and laissez-
faire physical to human capital ratio, the relative importance of factor accumulation may
switch, and we can formulate the following definition :
Definition 2 There is a relative factor accumulation (RFA) reversal, when the sign of the
term K = k∗ − kLF changes.
From equations (1.36) and (1.58), we have that K ≡ K (Υ). For simplicity’s sake, we
formulate this assumption, as relevant for a developed economy :
Assumption 5 α < 1/2.
Assumption 5 implies that Υ ∈ [−α ; α]. Using Lemma 6 and Definition 2, we com-
pare the optimum with the laissez-faire when sectoral factor intensity changes :
Proposition 5. Under Assumption 5, there exist three critical bounds β1, β2 and γ such
that :
(i) For δ > 1/2, then β2 < β1. If β > β1, kLF is always higher than k∗, if β2 < β < β1, ∃
Υ characterizing RFA reversal, and if β < β2, then kLF is always lower than k∗.
(ii) For δ < 1/2 and γ > γ, then β2 < β1 < β. If β1 < β, kLF is always higher than k∗, if
β2 < β < β1, ∃ Υ characterizing RFA reversal, if β < β2, kLF is always lower than k∗.
(iii) For δ < 1/2 and γ < γ, then β < β1 < β2. If β < β1, kLF is always lower than k∗, if
β1 < β < β2, ∃ Υ characterizing RFA reversal, if β2 < β, kLF is always higher than k∗.
Proof. see Appendix 1.10.4
Figure 1.1 illustrates Proposition 5, depicting zones of RFA reversal delimited by the
K curves. To give intuitions for Proposition 5, we observe the shape of the K curve, as
66
1.7 Laissez-faire and the social optimum
K > 0 (K < 0) means under (over) accumulation of physical capital. The occurrence of
a RFA reversal depends both on preferences (private and social) and technologies through
the capital share difference between sectors (Υ or ε = α0 − α1). Let us focus on this last
technological aspect.
(i) δ > 1/2 or (ii) δ < 1/2 and γ > γ (iii) δ > 1/2 and γ < γ
FIGURE 1.1 – Gap between the optimal and the laissez-faire physical to human ratio
Areas where RFA reversal can occur.
When the central planner has a high preference for future generations (case i), it allo-
cates more resources to produce the investment good than what agents would do (because
their future is restricted to one period ahead). The shape of the K curve then depends on the
capital share difference Υ. The optimal physical to human capital accumulation is higher
when the investment good is physical capital intensive (α0 < α1 and Υ < 0) than when
the consumption good is physical capital intensive (α0 > α1 and Υ > 0). This means that
when Υ increases, K goes up.
When the preference for future generations of the central planner is low (case iii), the
reverse results are obtained. The planner allocates more resources to produce the consump-
67
1 Optimal human and physical capital accumulation
tion good than what agents would do. Therefore, the optimal physical to human capital
accumulation increases compared to the competitive one (K goes up) when this good be-
comes more physical capital intensive (Υ goes up).
Finally, the remaining conditions in Proposition 5 simply prevents the RFA reversal to
occur, excluding cases where the K curve is always below or above the horizontal axis.
Thus, considering a two-sector model with human and physical capitals, we emphasize an
important result : when countries experience a factor intensity reversal, as was the case
in the postwar Japanese economy, the scheme of optimal capital accumulation may be
affected.
1.8 Policy implications
We examine how the government may proceed to decentralize the social optimum and
offset the market failure accruing from intergenerational externalities. From Docquier et.al
(2007), we know that in a one-sector framework with both human and physical capital ac-
cumulation, two instruments have to be implemented simultaneously to recover optimality.
This is because two sources of inefficiency are likely to interact in an OLG model with both
human and physical capital. This result is also true in our two-sector two-factor model, al-
though factor accumulation is not driven by the same mechanism. The aim of this section
is to highlight the effect of factor intensity ranking on the design of optimal policy.
From the previous section, we know that the size of each inefficiency is influenced by
sectoral properties. Using Proposition 5, we can formulate the following corollary, which
gives indications about optimal policy allowing to achieve the first-best optimum :
Corollary 1. (j) For δ > 1/2 or for δ < 1/2 and γ > γ : When β2 < β < β1, ∃ Υ, such
that if Υ > Υ (resp. Υ < Υ) it is efficient to invest relatively more (resp. less) in human
68
1.8 Policy implications
capital.
(jj) For δ < 1/2 and γ < γ : When β1 < β < β2, ∃ Υ, such that if Υ > Υ (resp. Υ < Υ )
it is efficient to invest relatively more (resp. less) in physical capital.
When the social planner mainly values future generation (case j), it is optimal to pro-
duce more in the investment sector. If Υ > Υ, the consumption good sector is physical
capital intensive, and more human capital is needed to produce more investment good.
Thus, the government has to devote more resources to education support.
For the design of optimal policy, we consider two fiscal instruments, as proposed in
Docquier et.al (2007). First, the standard intergenerational transfers between generations,
to correct failure in physical capital accumulation. Second, we consider an education sub-
sidy to achieve optimal human capital accumulation. Let θ be the education subsidy, τw
be the proportional tax on income and τ o be the lump sum tax on retirees. 19 Government
policy is the set (θ, τw, τ o). Government budget is balanced each period :
θtet = τwt wtht + τ ot ≡ θ(τw, τ o) (1.59)
Using competitive production conditions (1.28), (1.29), (1.30), and agent’s budget
constraints, we can rewrite the government budget constraint as : A1kα11t H1t = et + st.
This corresponds to the equilibrium on the investment good sector. From the FOC of the
agent’s program, we get :
1− θt =c∗th∗t+1
P ∗0t
(1− β)γb (A)
w∗
th∗
t (1− τwt ) = P ∗
0tc∗
t + s∗t + e∗t (1− θt) (B)
19. Notice that lump sum taxes during working period is strictly equivalent as long as endogenous laborsupply is not studied.
69
1 Optimal human and physical capital accumulation
R∗
t+1s∗
t − τ ot+1 = P ∗
0t+1d∗
t+1 (C)
Condition (A) gives the optimal education subsidy and ensures the optimal choice concer-
ning human capital investment. Conditions (B) and (C) give the optimal intergenerational
transfers such that optimality conditions are fulfilled. Because we consider a two-sector
model, the conditions (A), (B) and (C) allow to examine the scheme of public pension and
education subsidy with regards to the relative factor intensity differential. Using agent’s
optimal choices and the first-best solution, we can rewrite the conditions and deduce the
optimal policy scheme on the balanced growth path :
Proposition 6. Under Assumption 5, the optimal policy scheme is given by :
θ∗ =ψ(2Υ(δ − 1) + 1− α +Υ) + δγ(1− α +Υ)
ψ(2Υ(δ − 1) + 1− α +Υ) + γ(1− δ(α−Υ))(1.60)
τw∗ = 1− δ(1 + γ)(1− δ)
ψ(2Υ(δ − 1) + 1− α +Υ) + δγ(1− α +Υ)(1.61)
τ o∗ =δ(1 + γ(α−Υ))− ψ(2Υ(δ − 1) + 1− α +Υ)
γδ(1− α +Υ) + ψ(2Υ(δ − 1) + 1− α +Υ)(1.62)
With τ o∗ = τ o∗t+1/wt+1ht+1.
Proof. See Appendix 1.10.5.
From Lemma 4 and Proposition 6, we analyze the impact of factor intensity differences
across sectors on optimal policy, examining how an increase in Υ affects θ, τw∗ and τ o∗.
70
1.9 Conclusion
Corollary 2. Under Assumption 5, following an increase in Υ : When δ > 1/2, θ∗ and τw∗
increases while τ o∗ decreases. When δ < 1/2, there exists a β such that when β < β (resp.
β > β), θ∗ and τw∗ increases (resp. decreases), while τ o∗ decreases (resp. increases).
Proof. See Appendix 1.10.6.
We highlight that the sectoral gap shapes the design of optimal policy. When the plan-
ner’s discount rate is high and the investment good sector is human capital intensive, the
government has to devote more resource to education (θ high). Moreover, it has to transfer
income from the young saving generation to the old generation (τw∗ high and τ o∗ low)
in order to discourage physical capital accumulation. When the planner’s discount rate
is high and the investment good sector is physical capital intensive, the optimal policy
consists in accumulating more physical capital than human capital. The government has
to transfer income from the elderly people to the young saving generation and favor less
intensively education (θ and τw∗ low, and τ o∗ high). The reverse results are true when
planner’s discount rate and externalities in education are low (δ < 1/2 and β > β).
1.9 Conclusion
This paper shows how crucial it is to consider at least two-sector models to design op-
timal policy. Indeed, whereas an aggregated model hides the difference in factor intensities
across sectors, a multi-sector model allows such characteristics to be taken into account.
In this paper, we develop a two-sector model and we underline the importance of factor
intensity differences to design an optimal balanced growth path and optimal policy. We
conclude that changes in sectoral factor shares may imply a relative factor intensity re-
versal and thus affect the optimal accumulation of human and physical capital. A factor
71
1 Optimal human and physical capital accumulation
intensity differential between sectors should then be considered to determine the scheme
of optimal policy. We have shown that the two-sector model is tractable enough to conduct
such an analysis.
The trade-off between government support in physical or human capital depends on
market inefficiencies. The model we have studied is very stylized since we consider only
two keys inefficiencies identified in the literature : an intergenerational externalities in hu-
man capital and a specific inefficiency in physical capital arising from the OLG structure.
A natural extension of the model should introduce other externalities in human and phy-
sical capital. For example, education spending may generate positive effect on the total
factor productivity of each sector, while investment in physical capital may improve the
efficiency of education spending on human capital production, by allowing students to
have access to technical equipments more efficient. The inclusion of these features would
provide a more relevant analysis of the trade-off between physical and human capital. Re-
gardless of this, the main objective of this paper is to examine how sectoral properties
modify the optimal policy, and the simplified model that we consider already provides
interesting intuitions on this subject : the size of inefficiencies depends on the physical ca-
pital share in each sector. When sectors experience changes in factor intensity, the relative
importance of physical to human capital investment is modified.
72
1.10 Appendix
1.10 Appendix
1.10.1 Proof of Lemma 1
From the consumer budget constraints (a) and (b) and equations (1.33) and (1.34), we
have :
P0tct = wtht1− β
1 + γ
P0t+1dt+1 = Rt+1
(
β
1 + γwtht
)
Substituting these last expressions in the consumption goods market equilibrium (1.37)
gives :
1
(1 + γ)(wtht(1− β) +Rtβwt−1ht−1) = P0tA0k0t
α0hth0t
We divide this expression by ht and substitute (1.33) for ht−1 :
1
(1 + γ)
(
wt(1− β) +Rtβ
bγ(1 + γ)
)
= P0tA0k0tα0h0t (1.63)
As full employment of factors holds, we have k1h1 + k0h0 = k. From (1.27) this can be
written :
k1(1− h0) + k0h0 = k ⇒ h0 =k − k1k0 − k1
(1.64)
73
1 Optimal human and physical capital accumulation
Including (1.64) in equation (1.63) we obtain :
1
(1 + γ)
(
wt(1− β) +Rtβ
bγ(1 + γ)
)
= P0tA0kα00t
kt − k1tk0t − k1t
According to (1.36) for t > 0, kt+1 = kt = k. Using (1.30) and (1.36) :
1(1+γ)
(wt(1− β) +Rtk(1 + γ)) =
P0tA0
(
α0(1−α1)α1(1−α0)
)α0
Bα0(P0t)α0
α1−α0
(
k−k1t
B(P0t)1
α1−α0
(
α0−α1α1(1−α0)
)
)
fixing D ≡ (α0(1−α1))α0 (α1(1−α0))1−α0
α0−α1:
1
(1 + γ)(wt(1− β) +Rtk(1 + γa)) = P0tA0Dk1t
α0−1(k − k1t)
We replace P0t using (1.30) and factor returns by (1.28) and (1.29) :
1(1+γ)
(
(1− α1)A1k1tα1(1− β) + α1A1k1t
α1−1k(1 + γ))
= k1tα1−α0
Bα1−α0A0Dk1t
α0−1(k − k1t)
with DBα1−α0
= (α0(1−α1))α0 (α1(1−α0))1−α0
(α0−α1)
(
α1
α0
)α0(
A1
A0
)(
1−α0
1−α1
)α0−1
≡ (1−α1)α1
(α0−α1)A1
A0
As a result, the physical to human capital ratio in the investment good sector, k1, is constant
and we finally obtain, for t > 0 :
k1 = kα1
1− α1
(1− α1)− (α0 − α1)(
11+γ
)
(1− β)(α0 − α1) + α1
74
1.10 Appendix
To express the equilibrium growth rate, we use (1.38) with (1.32) and (1.33) :
1 + gt =γ
1 + γbAwt
As equilibrium physical to human capital ratio is constant at sectoral level, from (1.28),
the return of human capital is constant as well. We have wt = wt+1 = w, hence we obtain
a balanced growth path along which the variables chosen by agents (st, et, ct and dt+1)
grow at the same constant rate as individual human capital, g.
1.10.2 Proof of Lemma 2
Using equations (1.50) and (1.51) at time t+1, and the relationship h∗i = H∗i /H
∗ (with
i = 0, 1), gives :
A1k∗α11t+1H
∗
1t+1 −h∗t+2
b− k∗1t+2H
∗
1t+2 − k∗0t+2H∗
0t+2 = 0 (1.65)
and
A0k∗α00t+1H
∗
0t+1 − c∗t+1 − d∗t+1 = 0
Integrating (1.53) in the last equation, we can write :
H∗
0t+1 =c∗t+1
A0k∗α00t+1
ψ (1.66)
with ψ =(
(1−β)δ+β(1−β)δ
)
.
75
1 Optimal human and physical capital accumulation
Using FOC (1.45) and (1.47), we obtain the following relationship :
h∗t+1 =1
q∗1t
(
γbα1
α1 − (1− α1)k∗1t+1b
)
(1.67)
At each time, h = H1 +H0. Considering this relation, we can rewrite (1.65) :
A1k∗α11t+1(h
∗
t+1 −H∗
0t+1)−h∗t+2
b− k∗1t+2(h
∗
t+2 −H∗
0t+2)− k∗0t+2H∗
0t+2 = 0
and substitute h∗t+1 and h∗t+2 from equation (1.67) and H∗0t+1 and H∗
0t+2 from equation
(1.66). We obtain :
A1k∗α11t+1
[
1q∗1t
(
γbα1
α1−(1−α1)k∗1t+1b
)
− c∗t+1
A0k∗α00t+1
ψ]
= 1q∗1t+1
(
γα1
α1−(1−α1)k∗1t+2b
)
+k∗1t+2
[
1q∗1t+1
(
γbα1
α1−(1−α1)k∗1t+2b
)
− c∗t+2
A0k∗α00t+2
ψ]
+c∗t+2
A0k∗α00t+2
ψ
Simplify by q∗1t+1 and using equations (1.43) to (1.45) we have :
k∗1t+1
δ
(
γbα1−(1−α1)k∗1t+1b
)
− k∗1t+1
k∗0t+1(1− β)α0
α1ψ =
(
γα1
α1−(1−α1)k∗1t+2b
)
+ k∗1t+2
(
γbα1
α1−(1−α1)k∗1t+2b
)
− k∗1t+2
k∗0t+2δ(1− β)α0ψ + δ(1− β)α0ψ
(1.68)
From (1.44) to (1.47), optimal solution satisfies the equality of the marginal rate of trans-
formation between the two sectors :
(1− α1)A1k∗1t+1
α1
α1A1k∗1t+1α1−1 =
(1− α0)A0k∗0t+1
α0
α0A0k∗0t+1α0−1
76
1.10 Appendix
Thus, we obtain the following relationships between k∗0t+1 and k∗1t+1 :
k∗1t+1
k∗0t+1
=
(
1− α0
1− α1
)
α1
α0
Introduce it in (1.68) gives :
k∗1t+1
δ
(
γbα1−(1−α1)k∗1t+1b
)
−(
1−α0
1−α1
)
(1− β)ψ =
(
γα1
α1−(1−α1)k∗1t+2b
)
(
1 + k∗1t+2b)
−(
1−α0
1−α1
)
δ(1− β)α1ψ + δ(1− β)α0ψ(1.69)
We then compute the optimal physical to human capital ratio in the investment sector
along the BGP. From Definition 1, it is characterized by k∗1t+2 = k∗1t+1 = k1∗. We simplify
equation (1.69) and obtain :
γα1−(1−α1)k∗1b
(
k∗1b−α1δ−k∗1bδα1
δ
)
= δ(1− β)ψ(
α0 − α11−α0
1−α1
)
+ (1− β)ψ(
1−α0
1−α1
)
From (1.67), h∗ exists only when k∗1 > α1/b(1− α1). Therefore, we can write :
γk∗1b(1−δα1)−γδα1 =δ(1− β)ψ
1− α1
(δ (α0 − α1) + (1− α0))(α1−(1−α1)k∗
1b) (1.70)
Equation (1.70) leads to :
γ(1− δα1)k∗
1b+ Sk∗1b = γδα1 + S(
α1
1− α1
)
where S = δ(1− β)ψ(δ(α0 − α1) + (1− α0))
Therefore, we have an expression for k∗1 given by (1.55).
77
1 Optimal human and physical capital accumulation
From Definition 1 and (1.54) the optimal growth rate along the BGP is given by :
1 + g∗ = δα1A1k∗α1−11
This corresponds to the two-sector modified golden rule as 1 + g∗ = δf ′(k1).
Using equation (1.55) we compute the following derivatives :
∂k∗1∂γ
=δ − 1
δb(1− α1)2S
(γ(1− δα1) + S)2 < 0 ⇒ ∂g∗
∂γ> 0
∂k∗1∂β
=α1γ(1− δ)2
b
(1− α0) + δ(α0 − α1)
(γ(1− δα1) + S)2 > 0 ⇒ ∂g∗
∂β< 0
∂k∗1∂ε
= − α1
b(1− α1)
ψγ(1− δ)2
(S + γ(1− δα1))2< 0 ⇒ ∂g∗
∂ε> 0
Lemma 2 follows.
1.10.3 Proof of Proposition 4
We establish equation (1.58) from (1.57), substituting ε = 2Υ and α1 = α − Υ. The
derivative with respect to Υ is :
∂k∗
∂Υ=
−(ψ + γ)(ψ(2δ − 1) + δγ)
(ψΥ(2δ − 1) + ψ(1− α) + γ(1− δα) + δγΥ)2
The sign of this derivative is given by the term -(ψ(2δ−1)+ δγ). Including the expression
of ψ, ∂k∗
∂Υ> 0 when β(2δ− 1)(δ− 1)− δ((2δ− 1) + γ) > 0. We deduce the results given
in Proposition 4.
78
1.10 Appendix
1.10.4 Proof of Proposition 5
We compare the optimal and the laissez-faire physical to human capital ratio. The
laissez-faire ratio is given by (1.35) and the optimal by (1.58). From Defintion 2, when
K(Υ) = 0 the optimal and the laissez-faire physical to human capital ratio are equal. We
examine K(Υ) at the limits of its definition set : K(−α) = 2δαb(1−2δα)
− kLF and K(α) =
2(β(1−δ)+δ)(1−δ)αb[γ+(2α(δ−1)+1)(β(1−δ)+δ)]
−kLF , with K(−α) and K(α) decreasing in β. We have K(−α) >
0 if and only if β < γ2δα(1−2δα)
≡ β1 and K(α) > 0 if and only if :
R(β) ≡ −β2(1−δ)(2α(δ−1)+1)+β[
γ(2α(1− δ)2 − 1) + δ(2α(1− δ)− 1)]
+2γδα(1−δ) > 0
R(β) = 0 admits a unique positive solution β2 such that K(α) > 0 if and only if β < β2,
with :
β2 =(1− δ)2α((1− δ)γ + δ)− δ − γ +
√∆
2(1− α)(1 + 2α(δ − 1))
where ∆ = (1− δ)2α((1− δ)γ + δ)− δ − γ − 4(1− δ)2δγα(1− 2α(1− δ)).
According to Lemma 4, two cases are distinguished :
— When δ > 1/2, k∗ is decreasing in Υ and hence β2 < β1. When β > β1, K < 0 ∀
Υ. Conversely when β < β2, K > 0 ∀ Υ. When β2 < β < β1, there exists a critical
level Υ such that K(Υ) = 0, with :
Υ =ψβ(α− 1) + βγ(δα− 1) + αγ(ψ + δγ)
2ψδ(β + γ)− ψ(β + γ) + δγ(β + γ)
— When δ < 1/2, the value of β compared to β is decisive. When β = β we have
79
1 Optimal human and physical capital accumulation
K = (1−2δ)(1+2γ(1−δ)α)−γb(1−2δα)γ(1−2δ)(1−δ)
that is positive for γ < 1−2δ1−2(1−δ)α(1−2δ)
≡ γ and negative
for γ > γ.
— For γ > γ : β is higher than β1 and β2. If β > β, K < 0 ∀ Υ. If β < β, k∗ is
decreasing in Υ and we obtain the same result than when δ > 1/2.
— For γ < γ : β is lower than β1 and β2. If β 6 β, K > 0 ∀ Υ. If β > β, k∗
is increasing in Υ and hence β1 < β2. When β > β2, K < 0 ∀ Υ. Conversely
when β < β1, K > 0 ∀ Υ. When β1 < β < β2, there exists a critical level Υ
such that K(Υ) = 0.
Proposition 5 follows.
1.10.5 Proof of Proposition 6
From the planner’s FOC (1.46) and (1.47), we obtain the optimal choice concerning
human capital accumulation :
c∗t+1
h∗t+1
=
(
α1 − b(1− α1)k∗1t+1
bγα1
)
δ(1− β)A0α0k∗
0t+1α0−1
Using this expression with condition (A), (B), (C) and equations (1.28), (1.30), (1.33),
(1.34) and (1.53), we obtain :
θ =b(1− α1)k1
∗
α1
τw =bγk∗1(1− α1)− (α1 − b(1− α1)k
∗1)δ(1 + γ)
bγk∗1(1− α1)
τ o =
[
k∗ − β
(
α1 − b(1− α1)k∗1
bγα1
)]
α1
(1− α1)k∗1
80
1.10 Appendix
With τ o = τ ot+1/wt+1ht+1. We replace k∗1 and k∗ using (1.55) and (1.58) to finally obtain
expressions given in Proposition 6.
1.10.6 Proof of Corollary 2
Deriving equations (1.60), (1.61) and (1.62) with respect to Υ, we get :
sgn
(
∂θ∗
∂Υ
)
= sgn
(
∂τw∗
∂Υ
)
= sgn
(
−∂τo∗
∂Υ
)
≡ (2δ − 1)((1− β)δ + β) + δγ
When 2δ − 1 > 0, we have ∂θ∗
∂Υ> 0, ∂τw∗
∂Υ> 0 and ∂τo∗
∂Υ< 0. When δ < 1/2, the same
result is observed as long as β < δ(2δ − 1 + γ)/(1− 2δ)(1− δ) ≡ β.
81
Chapter 2
Public education spending, sectoral taxation, and growth
2.1 Introduction
When should a government allocate more resources to education ? The recent empirical
studies confirm the positive effect of public education spending on growth (De la Fuente
and Doménech, 2006, Cohen and Soto, 2007). From a theoretical view point, the direct
positive influence of public education on human capital accumulation is largely admitted.
Nonetheless, taking into account the negative impact of policy arising from its funding, the
relationship between growth and public education spending is generally non-monotonous.
Indeed, public education favors growth while the requisite taxation may depress it. The
contribution of this paper is to explore this link in a two-sector overlapping generation
growth model. The disaggregation of production into a manufacturing and a service sector
allows us to assess the growth implications of new tax schemes to finance public education
expenditure, namely sectoral taxes.
Since the emergence of the new growth theory initiated by Lucas (1988), human capital
accumulation has been identified as a major determinant of long-term economic growth. A
large part of the literature has considered the link between the level of public expenditure
on education and economic growth (Glomm and Ravikumar, 1992; 1997; 1998). In more
recent studies, authors have identified factors influencing this relationship. Blankenau and
82
2.1 Introduction
Simpson (2004) and Blankenau et al. (2007) emphasize that the effect of government spen-
ding on education depends on the level of government spending, the tax structure and the
production technologies. They found that the response of growth to public education ex-
penditure may be non-monotonic. Basu and Bhattarai (2012) emphasize that the elasticity
of human capital to public education is decisive. When this elasticity is high, countries
with a greater government involvement in education experience lower growth. None of
these models highlights a link between growth-enhancing policy and agents’ preferences.
This paper introduces a model with two goods : one sector produces a manufactured
good that can be either consumed or invested in physical capital, and the other produces
a service good that can be either consumed or invested in human capital. The government
allocates a fixed share of GDP to public education by levying a tax on manufacturing out-
put, on services, or on aggregate production. We reveal that public education expenditure
can enhance or reduce growth. It always improves directly human capital accumulation but
may affect negatively both private education spending and investment in physical capital.
The magnitudes of these two opposite effects are highly conditional on agents’ tastes as
long as taxation differs across sectors. In this case, agents’ preferences for time, human ca-
pital, and services shape the effect of public policy. Moreover, since education expenditure
is considered as a service, sectoral taxation changes the relative price of education. A tax
on the manufacturing sector favors education relative to physical investment, whereas a tax
on services makes the manufactured good more attractive. This relative price adjustment
reduces or reinforces the benefits of public education policies. From these properties, we
show that when public education is financed by a tax on manufacturing output, the share
of GDP allocated to education that maximizes the growth rate is higher in economies cha-
racterized by low preferences for services relative to manufactured goods. According to
83
2 Public education spending, sectoral taxation, and growth
the literature on growth and development, this characteristic is more often observed in de-
veloping economies (see e.g. Hseih and Klenow, 2008). The opposite result is obtained
when policy is financed by a tax on services.
In addition, we prove that sectoral taxes may perform better in terms of long-term
growth than a standard aggregate production tax. To finance public education, a tax on
manufacturing output is better for growth than a tax on aggregate production only if the
taste for services is low enough. Indeed, even if this funding reduces the relative price of
education, a low tax rate is required to guarantee that the crowding out of physical capital
is not too important. Conversely, a tax on the service sector performs better only when the
taste for services is sufficiently high. In this way, the negative impact of the policy, coming
from the increase in the education cost, is lower than with an aggregate tax.
The paper is organized as follows. In Section 2.2, we set up the theoretical model. Sec-
tion 2.3 is devoted to the impact of sectoral taxation regarding the relationship between
growth rate and public education. In Section 2.4, we compare the long-term growth rates
under the different funding systems. Section 2.5 provides an extension and Section 2.6
concludes.
2.2 The Model
We develop a two-sector overlapping generations model in which individuals live for
three periods. All individuals are identical within each generation and we assume there is
no population growth. Population size is normalized to unity. The initial adult is endowed
with K0 units of physical capital and H0 units of human capital.
84
2.2 The Model
2.2.1 Production technologies
We use the two-sector production structure proposed by Erosa et al. (2010). The repre-
sentative firm produces in the manufacturing (YM) and the services (YS) sector. Production
in both sectors results from Cobb-Douglas production technologies, using two inputs, hu-
man capital or effective labor supply H , and physical capital K. Let Ki and Hi, i =M,S,
be respectively the quantities of capital and effective labor used by sector i, production is
given by :
YMt = AMKαMtH
1−αMt AM > 0 (2.1)
YSt = ASKαStH
1−αSt AS > 0 (2.2)
with α ∈ (0, 1), the elasticity of output with respect to capital, which is assumed equal
across sectors. Manufacturing output can be either consumed or invested in physical capi-
tal while services are consumed or invested in human capital. Physical capital investment is
only private whereas human capital investment results from both public and private invest-
ment. Since in the OECD countries, on average, 90% of the current expenditure on public
education is devoted to teacher salaries 20, we consider that educational expenditures in
terms of services are empirically relevant. Both inputs are perfectly mobile between the
two sectors provided that :
HM +HS ≤ H, KM +KS ≤ K (2.3)
20. See OECD Indicator B6 : “On what resources and services is education funding spent ?", available athttp ://www.oecd.org/education/eag.htm
85
2 Public education spending, sectoral taxation, and growth
K being the total stock of physical capital and H the total amount of human capital.
Let ki = Ki/Hi be the physical to human capital ratio of sector i, hi = Hi/H be the
share of human capital allocated to sector i, i = M,S, and k = K/H the physical to
human capital ratio. Equations (2.1), (2.2) and (2.3) can be rewritten :
YMt = AMkαMthMtHt (2.4)
YSt = ASkαSthStHt (2.5)
hM + hS ≤ 1, kMhM + kShS ≤ k (2.6)
The government collects revenue through a sector specific tax on output τi ∈ [0, 1), i =
M,S. We normalize the price of manufactured good to one. Denoting by w the wage rate,
R the gross rental rate of capital and PS the price of services, profit maximization over the
two sectors implies that production factors are paid at their net-of-tax marginal product :
Rt = (1− τMt)AMkMα−1t = (1− τSt)PStASkS
α−1t (2.7)
wt = (1− τMt)AMkMαt = (1− τSt)PStASkS
αt (2.8)
From which we have :
kMt = kSt = kt ; PSt =AM (1−τMt)AS(1−τSt)
(2.9)
Equation (2.9) shows that the relative price of services increases with the tax on services
whereas it decreases with the tax on manufacturing output.
We assume that physical capital fully depreciates after one period. In line with Blan-
86
2.2 The Model
kenau and Simpson (2004), the human capital accumulation is given by :
ht+1 = AHeat v
bth
1−a−bt a, b ∈ [0, 1], AH > 0 (2.10)
Parameters a and b are respectively the elasticity of human capital to private (et) and public
education (vt) expenditure. Public and private education expenditures are imperfect sub-
stitutes in producing human capital. In line with Keane and Wolpin (2001), et represents
resources that households invest in their children outside school (individual teachers, tui-
tions payments...). To keep the impact of the stock of parental knowledge (ht) on children’s
human capital positive, we restrict a+ b < 1.
2.2.2 Government
We assume that a fixed share (θ) of GDP (Yt) is devoted to public education, i.e PStvt =
θYt where Yt = YMt + PStYSt. From equations (2.4) and (2.5), the public expenditure on
education is :
PStvt = θkαt Ht (AMhMt + AShStPSt) (2.11)
Production taxes supported by the firms are the only source of government income. Go-
vernment policy is the set τS, τM , θ and government budget constraint is given by :
PStvt = τStPStYSt + τMtYMt (2.12)
Using (2.4), (2.5), (2.9) and (2.11), the government budget constraint can be written :
θ (hMt(1− τSt) + hSt(1− τMt)) = τSthSt(1− τMt) + τMthMt(1− τSt) (2.13)
87
2 Public education spending, sectoral taxation, and growth
2.2.3 Preferences
The economy is populated by finite-lived agents living for three periods. We consider
a paternalistic altruism, according to which parents value the level of human capital of
their children. In their first period of life, agents are young and benefit from education. In
their second period of life, agents are adult and they are endowed with ht efficiency units
of labor that they supply inelastically to firms. Their income is allocated between current
consumption, ct, savings, st and investment in children’s education, et.
wtht = πtct + PStet + st (2.14)
In their third period of life, agents are old and retire. They consume the proceeds of their
savings :
Rt+1st = πt+1dt+1 (2.15)
We denote by π the price of the composite good c, which is an aggregate of the manu-
factured and the service goods. Let x = c, d denote the individual consumption at each
period of life, xM and xS be respectively the quantities allocated to manufactured goods
and services. Instantaneous preferences over the two goods are defined according to :
x = xµMx1−µS (2.16)
with µ ∈ (0, 1) the share of manufactured goods in consumption. The optimal allocation
of total expenditure between consumption of manufactured goods and services is obtained
88
2.2 The Model
by solving the following static problem :
maxxM ,xS
xµMx1−µS
s.t πx = PSxS + xM
and leads to :
xM = µπx ; PSxS = (1− µ)πx
π = φ (µ) ≡ µ−µ (1− µ)−(1−µ)
(2.17)
An individual born in period t−1 chooses et and st so as to maximize his life-cycle utility :
U(ct, dt+1, ht+1) = ln ct + β ln dt+1 + γ lnht+1 (2.18)
0 < β < 1 ; 0 < γ < 1
subject to (2.10), (2.14) and (2.15). Parameters β and γ are respectively the discount factor
and the degree of paternalistic altruism.
From the first order conditions, we obtain individual’s optimal choices :
st =β
1 + γa+ βwtht (2.19)
et =γa
PSt(1 + γa+ β)wtht (2.20)
89
2 Public education spending, sectoral taxation, and growth
Since education expenditure is assumed to be in terms of service only, sectoral taxation
creates a distortion by affecting the price of education relative to saving.
2.2.4 Equilibrium
Definition 3. Given a set of initial conditions K0, H0, an equilibrium is a sequence of
prices wt, Rt, PStt=∞t=0 , decision rules cMt, cSt, dMt+1, dSt+1, st, et, ht+1t=∞
t=0 and quan-
tities Kt, ht, Ytt=∞t=0 such that, for all t > 0 :
i) A period t adult chooses cMt, cSt, dMt+1, dSt+1, st, et, ht+1 to solve the agent’s pro-
blem taking prices and government policy as given ;
ii) (wt, Rt, PSt) is given by (3.7) and (3.8) ;
iii) the effective labor supply in t is Ht = ht ;
iv) the service goods market clears : YSt = cSt + dSt + et + vt ;
v) the physical capital market clears : Kt+1 = st ;
vi) the budget constraint clears : θ(YMt + PStYSt) = τStPStYSt + τMtYMt ;
The clearance of the goods markets in period t requires the demand for services (i.e.,
the sum of consumption of service goods and public and private spending on education)
to be equal to the supply of the service goods :
Lemma 4. The clearance of the service goods market in period t
YSt = cSt + dSt + et + vt (2.21)
gives the share of human capital allocated to each sector :
hSt = X + τMt(1−X ) ; hMt = (1− τMt)(1−X ) (2.22)
90
2.2 The Model
with X =(1− α)γa+ (1− µ)(1 + α(γa+ β))
1 + γa+ β< 1 (2.23)
Proof. See Appendix 2.7.1
By substituting equation (2.22) to (2.13) we deduce the relationship between θ, τM and
τS :
θ =τStX + τMt(1−X )
1 + (τMt − τSt)(1−X )(2.24)
We study successively a tax on manufacturing (τS = 0) and services production (τM = 0).
As a result, a constant share θ means that tax rates are time invariant. The capital market
clearing condition with equation (2.10) gives :
kt+1 =st
AHeat vbth
1−a−bt
Using equations (2.11), (2.19), (2.20) and (2.22) we finally obtain the dynamic equation
characterizing equilibrium paths :
kt+1 =AMβ(1− τM)(1− α)1−ak
α(1−a−b)t
AH(γa)aAa+bS (1− τS)a(1 + γa+ β)1−aθb(1 + (1−X )(τM − τS))b
(2.25)
The dynamic path of kt is monotonic and converges toward the following steady state
value :
k =
(
AMβ(1− τM)(1− α)1−a
AH(γa)aAa+bS (1− τS)a(1 + γa+ β)1−aθb(1 + (1−X )(τM − τS))b
)1
1−α(1−a−b)
91
2 Public education spending, sectoral taxation, and growth
Then, we obtain a balanced growth path equilibrium along which the variables chosen
by individuals (st, et, ct and dt+1) and public education expenditure (vt) grow at the same
constant rate as human capital :
1 + g =ht+1
ht= AH
(
AS kα)a+b
(
γa(1− α)(1− τS)
1 + γa+ β
)a
θb(1 + (1−X )(τM − τS))b
In the following, we focus on the balanced growth path.
2.3 Public education funding and long-term growth rate
We examine the relationship between public education expenditure and long-term
growth considering different types of funding. The decomposition of the aggregate eco-
nomy into two sectors allows us to consider sectoral taxation. We define εij as the elasticity
of iwith respect to j and z as the private education spending per unit of human capital e/h.
We examine three alternative policies to finance public education.
2.3.1 Public education financed by a tax on aggregate production
Assume that τM = τS = τY , the fiscal policy is equivalent to a tax on aggregate
production. Factor returns given by equations (2.7) and (2.8) are negatively affected by the
tax, whereas the relative price of goods remains unchanged. From equation (2.24), the tax
rate is equal to the share of output devoted to education spendings :
τY = θ
92
2.3 Public education funding and long-term growth rate
The physical to human capital ratio, private education per unit of human capital, and long-
term growth rate are respectively given by :
k =
(
AMβ (1− θ)1−a (1− α)1−a
AH(γa)aAa+bS θb(1 + γa+ β)1−a
) 11−α(1−a−b)
z =γa
1 + γaAS(1− α)kα(1− θ)
g = AH
(
AS kα)a+b
(
γa(1− α)
1 + γa+ β
)a
θb(1− θ)a (2.26)
The impact of an increase in the share of GDP devoted to public education on private
choices and growth is deduced from the elasticities :
Lemma 5. When the government taxes the aggregate production, the elasticities are given
by :
εk,θ = − 1
1− α + α(a+ b)
(
(1− a)θ
1− θ+ b
)
< 0
εz,θ = αεk,θ − 1 < 0
εg,θ = α(a+ b)εk,θ + b− aθ
1− θ≶ 0
Proof. See Appendix 2.7.2
We obtain results similar to Blankenau and Simpson (2004), the differences being due
to our formalization of agents’ preferences for education and the scheme of public po-
licy. 21 Elasticities do not depend on agents’ preferences and an increase in public educa-
21. In Blankenau and Simpson (2004), agents borrow for education when young and the governmentallocates a share of its budget to unproductive spendings.
93
2 Public education spending, sectoral taxation, and growth
tion spendings crowds out both physical capital accumulation and private human capital
investment. Thus, there is a growth-maximizing level of public expenditures :
Proposition 7. When the government taxes the whole production at the same rate, the level
of public expenditure maximizing the growth rate is given by :
θmaxY =
a+ b
(1− α)a
Policy is growth-enhancing (resp. reducing) when θ < θmaxY (resp. θ > θmax
Y ).
The relationship between growth and public education spending does not depend on
agents’ preferences as long as the level of tax is the same in both sectors.
2.3.2 Public education financed by a tax on manufacturing output
Assume that τS = 0. We focus on the growth effect of public education spending on
the long-term growth rate when public intervention is financed by a tax on the production
of manufactured goods only. This positive tax causes a fall in factor returns. Moreover,
it creates a distortion making education more attractive. Indeed, education and services
become cheaper relative to manufactured goods. From equation (2.24), a balanced budget
constraint requires :
τM =θ
(1− θ)(1−X )
Policy is sustainable (τM < 1) if θ is not too high : θ < 1−X
2−X≡ θ. A higher θ is associated
with a lower share of human capital allocated to the production of manufacturing output
and a higher tax on this output. With X given by (2.23), examining the expression of τM
we emphasize the following properties :
94
2.3 Public education funding and long-term growth rate
Proposition 8. The tax rate on manufacturing production required to balance the public
budget is increasing with altruism factor (γ) and decreasing with time preference (β) and
taste for manufactured goods (µ).
An economy characterized by a low taste for education (γ), a high degree of time pre-
ferences (β) and a high taste for manufactured goods (µ), will be oriented toward the
consumption of manufactured goods. Since the government taxes this sector to finance
public education expenditure, the required tax rate will be low.
The physical to human capital ratio, private education per unit of human capital, and
long-term growth rate are respectively given by the expressions :
k =
AMβ(
(1−θ)(1−X )−θ(1−θ)(1−X )
)
(1− α)1−a
AH(γa)aAa+bS (1 + γa+ β)1−a
(
θ1−θ
)b
11−α(1−a−b)
z =γa
1 + γaAS(1− α)kα
g = AH
(
AS kα)a+b
(
γa(1− α)
1 + γa+ β
)a(θ
1− θ
)b
(2.27)
Thus, we compute the following elasticities :
Lemma 6. When public policy is financed by a tax on manufacturing production, the elas-
ticities are given by :
εk,θ = − 1
(1− θ)(1− α + α(a+ b))
(
θ
(1− θ)(1−X )− θ+ b
)
< 0
εz,θ = αεk,θ < 0
95
2 Public education spending, sectoral taxation, and growth
εg,θ = α(a+ b)εk,θ +b
1− θ≶ 0
Proof. See Appendix 2.7.2
As in the case where sectors are taxed at the same rate, an increase in public education
spending crowds out investment in physical capital because taxation reduces wage, and
therefore, the amount of saving. Regarding private education choices, the ratio w/PS is
a crucial variable. The tax decreases the price of education, thus public intervention in
education affects this ratio only through the modification of physical capital intensity (k).
Consequently, a tax on the manufacturing sector reduces the crowding out effect of policy
on private education choices. Using εg,θ and εk,θ we easily see that policy has a non-
monotonic impact on the growth rate which crucially depends on the agents’ preferences.
Proposition 9. Under manufacturing-tax funding system, the policy maximizing the long-
term growth rate is :
θmaxM =
b(1− α)(1−X )
b(1− α)(1−X ) + aα + b< θ ; τmax
M =b(1− α)
aα + b
Policy is growth-enhancing (resp. reducing) when θ < θmaxM (resp. θ > θmax
M ).
The elasticity of the growth rate to public education (εg,θ) is decreasing with prefe-
rences for education (γ) and the share of services in total consumption (1 − µ). This is
directly linked to the higher level of output taxation required when economy is orien-
ted toward services. Countries with a high preference for manufactured goods experience
higher growth rate when the tax concerns manufacturing production. Therefore, policy
recommendations are taste-dependent.
96
2.3 Public education funding and long-term growth rate
Corollary 3. In the manufacturing-tax funding system, the higher the consumption taste
for manufactured goods relative to services, the higher the share of GDP devoted to edu-
cation that maximizes the growth rate.
2.3.3 Public education financed by a tax on services
We assume now the case where τM = 0, that is public intervention is exclusively
financed through a tax on services. From the firm’s optimization program, taxation of
services differs from taxation of the manufacturing sector in two ways. A tax on services
does not affect the factor return, however, it raises the price of education. From (2.24), the
tax rate balancing the budget constraint is the following :
τS =θ
X + (1−X )θ
Proposition 10. The tax rate on services required to balance public budget is decreasing
with altruism factor (γ) and increasing with time preference (β) and taste for manufactured
goods (µ).
An economy characterized by a high taste for education (γ), a low degree of time
preferences (β) and a low share of manufactured goods in consumption expenditure (µ),
will be oriented toward the consumption of services. A high demand for services entails a
large scale of the production of this good. In a services-tax funding system this guarantees
that the tax rate is not too high. As previously, we compute the physical to human capital
97
2 Public education spending, sectoral taxation, and growth
ratio, private education per unit of human capital, and long-term growth rate :
k =
AMβ(1− α)1−a
AH(γa)aAa+bS
(
X (1−θ)X+(1−X )θ
)a
(1 + γa+ β)1−a(
θXX+(1−X )θ
)b
11−α(1−a−b)
z =γa
1 + γaAS(1− α)kα
( X (1− θ)
X + (1−X )θ
)
g = AH
(
AS kα)a+b
(
γa(1− α)
1 + γa+ β
)a( X (1− θ)
X + (1−X )θ
)a(θX
X + (1−X )θ
)b
(2.28)
The elasticities are derived in the following Lemma :
Lemma 7. When public education expenditure is financed by a tax on services, elasticities
are given by :
εk,θ =aθ − b(1− θ)X
(1− θ)(1− α + α(a+ b))(X + (1−X )θ)≶ 0
εz,θ = αεk,θ −1
(1− θ)(X + (1−X )θ)< 0
εg,θ = α(a+ b)εk,θ +X b(1− θ)− aθ
(1− θ)(X + (1−X )θ)≶ 0
Proof. See Appendix 2.7.2
In contrast to the case where policy is financed by a manufacturing output tax, an
increase in public education expenditures does not always reduce the physical to human
capital ratio. The introduction of a tax on the production of services increases the price
of services (and the price of education) by the full amount of the tax. Thus, it generates
98
2.3 Public education funding and long-term growth rate
opposite effects on the physical to human capital ratio : a negative effect coming from
the increase in public spending on education and a positive effect arising because the
education expenditure becomes more expensive compare to investment in physical capital.
The positive effect dominates when the economy is more oriented toward manufactured
goods than services (X low). The private education choice per unit of human capital goes
down when the government increases public education (εz,θ < 0). Even when policy favors
the return of human capital through the raise in the physical to human capital ratio (εκ,θ >
0), the negative effect generated by the increase in the relative price is higher. The global
impact of a raise in θ on the long-term growth rate is given by εg,θ. It is ambiguous and
depends on agents’ preferences :
Proposition 11. Under service-tax funding system, the policy maximizing the long-term
growth rate is :
θmaxS =
bXa+ bX ; τmax
S =b
a+ b
Policy is growth-enhancing (resp. reducing) when θ < θmaxS (resp. θ > θmax
S ).
The elasticity of the growth rate to public education (εg,θ) is increasing with agents’
taste for education and the share of services in total consumption. 22 As previously, this is
because a low level of output taxation is required when economy is service sector oriented.
Thus, we emphasize the following result :
Corollary 4. In the service-tax funding system, the higher the preferences for services
relative to manufactured goods, the higher the share of GDP devoted to education that
maximizes the growth rate.
22. Note that θmaxS does not depend on the elasticity of output to physical capital (α). The direct impact
of policy on the long-term growth rate, captured by the second term of the right hand side of εg,θ, alwaysneutralizes the indirect impact generated by the adjustment of k.
99
2 Public education spending, sectoral taxation, and growth
In this model, we highlight that government has to adjust its policy according to the
pattern of consumption to achieve the higher growth rate. Based on the literature on growth
and development, it appears that rich countries are more oriented toward services than de-
veloping countries (see e.g Hseih and Klenow (2008)). As a result, taking into account
sectoral taxation, we can conclude that the relationship between growth and public educa-
tion expenditure is not the same along the process of development. Conversely, with a tax
on aggregate production the relationship between growth and public education would not
depend on agents’ tastes.
2.4 Sectoral tax versus aggregate output tax
We have shown in the previous section that the relationship between growth and public
education spending is not monotonous. Moreover, when sectoral taxes finance public edu-
cation, the design of a growth-enhancing policy depends on preferences and policy shapes
the long-term growth through an additional channel. Sectoral taxes create a distortion by
changing the relative price of education, which amplifies or weakens the positive effect of
policy.
We compare the long-term growth rate in different fiscal regimes presented previously.
More precisely, we examine weather a distortionary sectoral production tax has to be pre-
ferred to a tax on aggregate production to finance public education expenditure. We define
by gM , gS and gY the growth rates with manufacturing, services and aggregate production
taxes respectively. The trade-off depends on the share of manufactured goods in consump-
tion expenditure (µ) :
100
2.4 Sectoral tax versus aggregate output tax
Proposition 12. For a given θ,
j) There exists a critical level µM such that : when µ > µM (resp. µ < µM ), gM > gY
(resp. gM < gY ).
jj) There exists a critical level µS such that : when µ < µS (resp. µ > µS), gS > gY
(resp. gS < gY ).
with critical values µM and µS decreasing in β and increasing in γ. 23
Proof. See Appendix 2.7.3
The long-term growth rate is higher when public education policy is financed by a tax
on the service sector rather than a tax on the aggregate production, as long as the consump-
tion of services is important (µ low), the taste for children education is high (γ high) and
the time preference is low enough (β low). When government taxes the production of ser-
vices only rather than the aggregate production, two additional opposite effects arise. On
the one hand, the factor returns are not directly affected by taxation, making the return
of human capital higher. On the other hand, education becomes more expensive. With a
high taste for services, taxation is not too high. In this case, the positive effect dominates
and allows to achieve a higher growth rate. Concerning the comparison between aggregate
taxation and a tax on manufacturing output a symmetric result emerges. The wage and
the price of education are lower with a tax on manufacturing sector than with an aggre-
gate one. Consequently, the reduction of physical to human capital ratio is more severe,
whereas the opposite result holds for private education spending. Financing an increase of
public education policy by imposing a tax on the manufacturing output performs better,
provided that the tax is not too high. It is the case when consumption of manufactured
23. Expression for µM and µS are given in Appendix.
101
2 Public education spending, sectoral taxation, and growth
goods is sufficiently important (µ high, γ low and β high). 24
2.5 Extension : factor intensity differential between sectors
In our model, the factor intensity differential between sectors has been neglected in
order to obtain a sharp characterization of the results. Thus, the relative price depends
only on total factor productivity and fiscal policy gap between sectors. It follows that the
relationship between growth and public education expenditure is affected by preferences
only when the tax rate that balances the government budget is taste-dependent. This result
does not hold when factor intensities differ across sectors because the relative price is then
determined by the equilibrium on the service goods market. Assuming that the service
sector is more human capital intensive than the manufacturing one we have :
YMt = AMkαMMt hMtHt
YSt = ASkαSSt hStHt
where αS < αM , and from the profit maximisation :
kMt = B(PSt)1
αM−αS
kSt =αS(1−αM )αM (1−αS)
B(PSt)1
αM−αS
with B =(
αS
αM
)
αSαM−αS
(
AS
AM
) 1αM−αS
(
1−αM
1−αS
)
αS−1
αM−αS
In this case, the relative price adjusts to the equilibrium on the goods market and an addi-
tional mechanism emerges : A raise in θ increases the relative price of goods by favoring
24. Regarding threshold levels µM and µS , a situation where both kinds of distortionary sectoral taxesperform better than the tax on aggregate production (µM < µ < µM ) is not excluded. Nevertheless, analy-tically, we can not conclude in favor of one type of sectoral taxation or the other.
102
2.6 Concluding remarks
the consumption of services. This raise has two consequences. On the one hand, it makes
education spending more costly. On the other hand, from the Stopler Samuelson theorem,
it leads to an increase in wage. These two effects impact the growth rate in opposite direc-
tions but would not alter the main conclusions obtained when public education is financed
by distortionary sectoral taxes. When manufacturing sector is taxed, the direct impacts of
policy overtake these indirect effects such that agents’ preferences influence the relation-
ship between growth and public education in the same way. When services production
is taxed, these effects are exactly offset by the distortionary impact of sectoral taxation,
which reduces the amount of education spending.
However, as long as the service sector is more human capital intensive than the ma-
nufacturing one, the result according to which the growth effects of a policy financed by
an aggregate tax do not depend on agents’ preferences is no longer satisfied. Indeed, even
if the amount of tax remains not taste-dependent, the modification of the relative and its
consequences are influenced by agents’ tastes. Nevertheless, since these effects work in
the opposite direction, the impact of preferences on the relationship between growth and
public education expenditure stays ambiguous. It depends on technology parameters in
human capital accumulation and production functions.
2.6 Concluding remarks
The effects of a public education policy financed by sectoral taxes differ from those
obtained in a one-sector model where the policy is financed by a standard production tax.
Since education expenditures are assumed to be a service only, a sectoral tax creates a dis-
tortion by modifying the relative price of education. Moreover, it makes the agents’ tastes
a major determinant of the relationship between growth and public education expenditure.
103
2 Public education spending, sectoral taxation, and growth
Cross-country heterogeneity in preferences for human capital, services and savings deter-
mine the sizes of the manufacturing and service sectors and hence shapes the design of
a growth-enhancing education policy financed by distortionary sectoral taxes. The com-
parison between the different tax schemes allows us to show that the growth rate may be
higher when the policy is supported by a sectoral production tax rather than a tax on the
aggregate production.
We would like to acknowledge that our analysis abstracts from important features that
call for further research. In particular, the present framework underlines the importance of
the relative sizes of the manufacturing and service sectors in a country by considering a
closed economy. Using a two-sector two-country model should provide additional insights
on the implication of sectoral taxation on the relationship between growth and public edu-
cation expenditure.
2.7 Appendix
2.7.1 Proof of Lemma 4
From Eq. (2.14), (2.15), (2.17), (2.19) and (2.20) we have :
PStcSt =(1− µ)wtht1 + γa+ β
; PStdSt = (1− µ)st−1Rt
Using these expressions and Eq. (2.12) and (2.20), the clearance of the service good’s
market is :
PStYSt =(1− µ)wtht1 + γa+ β
+ (1− µ)st−1Rt +γawtht
1 + γa+ β+ τStYStPSt + τMtYMt
104
2.7 Appendix
Including (2.4), (2.5) and factor returns :
AM(1− τMt)kαt hStht =
AM (1−α)(1−τMt)kαt ht(1−µ+γa)
1+γa+β+ (1− µ)st−1AMα(1− τMt)k
α−1t + τMtAMk
αt hMtht
Simplifying and using the equilibrium on the physical capital market st−1ht = kt :
hSt =(1− α)(1− µ+ γa)
1 + γa+ β+ α(1− µ) +
τMt
1− τMt
hMt
As hTt = 1− hNt, we easily obtain Eq. (2.22).
2.7.2 Proof of Lemma 5, 6 and 7
Elasticities presented in Lemma 5, 6 and 7 are computed using derivatives ∂k∂θ
, ∂z∂θ
and
∂g∂θ
. We determine these derivatives for each regime.
Tax on aggregate production :
∂k
∂θ=
(
1− a
1− θ− b
θ
)
k
1− α + α(a+ b)
∂z
∂θ= α
∂k
∂θ
z
k− z
1− θ
∂g
∂θ= α(a+ b)
∂k
∂θ
g
k+ g
(
b
θ− a
1− θ
)
g
Manufacturing-tax funding system :
∂k
∂θ= −
(
1
(1− θ)(1−X )− θ+b
θ
)
k
(1− θ)(1− α + α(a+ b))
105
2 Public education spending, sectoral taxation, and growth
∂z
∂θ= α
∂k
∂θ
z
k
∂g
∂θ= α(a+ b)
∂k
∂θ
g
k+ g
b
(1− θ)θ
Service-tax funding system :
∂k
∂θ=
(
(a+ b)(1−X )
X + θ(1−X )− b
θ+
a
1− θ
)
k
1− α + α(a+ b)
∂z
∂θ= z
(
α
k
∂k
∂θ− 1
(1− θ)(X + (1−X )θ)
)
∂g
∂θ= α(a+ b)
∂k
∂θ
g
k− g
(
(a+ b)(1−X )
X + θ(1−X )− b
θ+
a
1− θ
)
2.7.3 Proof of Proposition 12
We give the condition which guarantees gM > gY , using Eq. (2.27) and (2.26) :
(
θ
1− θ
)b(1− α(a+b)1−α(1−a−b))((1− θ)(1−X )− θ
(1− θ)(1−X )
)α(a+b)
1−α(1−a−b)
> θb(1−θ)a(
(1− θ)1−a
θb
)
α(a+b)1−α(1−a−b)
After simplifications we obtain :
(1− θ)(1−X )− θ
(1− θ)(1−X )> (1− θ)
1+αα
Replacing expression X by Eq. (2.23), we finally get :
µ > 1−
(1− θ)(
1− (1− θ)1α
)
− θ
(1− θ)(
1− (1− θ)1α
)
1 + γa+ β
1 + α(γa+ β)+
(1− α)γa
1 + α(γa+ β)≡ µM
106
2.7 Appendix
Then, we determine the condition which guarantees gS > gY , using Eq. (2.28) and (2.26) :
[
( X (1− θ)
(1−X )θ + X
)a( X θ
(1−X )θ + X
)b]1−
α(a+b)1−α(1−a−b)
> θb(1− θ)a(
(1− θ)1−a
θb
)
α(a+b)1−α(1−a−b)
After simplifications, we obtain :
XX + (1−X )θ
> (1− θ)α
1−α
and with Eq. (2.23), we get :
µ < 1− (1− θ)α
1−α θ
1− (1− θ)α
1−α
1 + γa+ β
1 + α(γa+ β)+
(1− α)γa
1 + α(γa+ β)≡ µS
107
Chapter 3
Short-and long-term growth effects of integration in
two-sector economies with non-tradable goods
3.1 Introduction
The process of economic integration is a major challenge especially in Europe. Since
2004, European Union (EU) membership has grown from 15 to 28 countries and the ana-
lysis of the economic benefits of such enlargement is a high issue in the political agenda.
As underlined by Kutan and Yigit (2007), integration shapes many aspects of an eco-
nomy, what make the evaluation of its overall impact difficult. There is no consensus yet
on this question, even in the empirical literature. From a theoretical point of view, static
implications of economic integration have been largely documented through the standard
trade theory. These analysis have been completed by studies examining the dynamical
consequences of international trade on growth (see for example Rivera-Batiz and Romer,
1991) 25. According to this literature, integration generates factor reallocations between
asymmetric countries that shape the growth rate.
More recently, some authors have enhanced this topic by considering education choices
and human capital accumulation. Michel and Vidal (2000) examines the long-term growth
effects of economic integration when patience and altruism drive respectively physical and
25. Rivera-Batiz and Romer (1991) focus on the pure scale effect of integration by considering trade bet-ween similar countries. They show that the increase in the flow of ideas, generated by integration, improvesthe productivity of research in both regions
108
3.1 Introduction
human capital accumulation. They obtain that two countries can benefit from integration
when cross-border externalities in human capital are high enough. Galor and Mountford
(2008) highlight the influence of international trade on human capital. They show that
trade increases education in OECD countries while it decreases it in non-OECD countries.
While it is widely admitted that education spending is a non-tradable good, these papers
do not consider the existence of non-tradable production. Moreover, the consumption of
non-traded goods represents a significant part of the aggregated consumption. Dotsey and
Duarte (2008) underline that consumption of non-traded good represents about 40% of US
GDP whereas Berka and Devereux (2013) claim that 30% of the aggregated consumption
is non-tradable for European countries. Thus, this paper aims at examining the short-and
long-term implications of economic integration when a part of production is non-tradable.
We extend the paper of Michel and Vidal (2000) by considering a two-sector mo-
del with human and physical capital accumulation. Economic integration results from the
perfect mobility of physical capital between two countries and generates cross-border ex-
ternalities in human capital. In the context of the Bologna Process, the mobility of students
across Europe steadily increases this last decade. The European Commission works clo-
sely with policy makers to promote mobility of students and cross-border cooperations.
Thus, we assume that human capital in the home country depends on education spending
in the foreign one. The decomposition of the aggregate economy into a tradable and a
non-tradable sector allows us to consider sectoral TFP disparities between countries and
sectoral factor share. Such distinction is empirically relevant : There exist TFP spreads
between countries, but these TFP spreads are sector-specific. For example, Hseih and Kle-
now (2008) show that TFP spreads are higher in the investment good sector. Herrendorf
and Valentinyi (2008) emphasize that the TFP spreads between developing countries and
109
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
the US are smaller in the non-tradable sector. They also show that sectoral factor shares
vary considerably across sectors 26. While Michel and Vidal (2000) performs only a long-
term analysis, our framework allows to investigate the short-run dynamics and to identify
new determinants of the growth effects of economic integration. The presence of the non-
tradable goods introduces a key variable for factor allocation in the economy : the price
of the non-tradable goods in terms of the traded goods. This relative price determines the
return of factor and so the growth rate.
Our study introduces a new dimension : the transitional dynamics in the integrated eco-
nomy is driven by both the dynamics of the relative price and the dynamics of foreign over
domestic education spending. When heterogeneous countries integrate, there is a transitio-
nal adjustment of the relative prices which affects the transitional dynamics of the growth
rate. Such adjustment depends mainly on the TFP spreads in the tradable sector. A high-
traded TFP country exhibits a high interest rate. Following integration, physical capital
goes from the low-traded TFP to the high-traded TFP country. When the non-tradable sec-
tor is human capital intensive - like it is the case in developed countries - this entails a fall
in the relative price of non-tradable good, and hence of the growth rate, for the low-traded
TFP. This effect is transitional and reduces across time.
We emphasize that the long-term consequences of integration on economic growth
does not result only from the differences of time preferences and education preferences
between countries. In Michel and Vidal (2000) the autarky growth rate of a low altruistic
and impatient country (i.e a country that invests less in human and physical capital compa-
red to other countries) is always lower than the growth rate in the integrated economy. This
result does not hold in our setting because the wage is determined by the relative price of
good that depends on the taste for the non-tradable good and the sectoral TFPs. When the
26. For example, food has a labor share of only 0.62 while construction has a labor share as high as 0.79.
110
3.2 The model
agents’ taste for tradable goods is higher in the domestic country, the autarky’s domestic
relative price of the non-tradable goods in terms of tradable is lower. When the tradable
sector is physical capital intensive the autarky’s domestic wage is lower and the autarky’s
domestic return on physical capital is higher than in the foreign country. Integration leads
to a convergence in return on capital and entails an increase (a decrease) in the wage for the
domestic (foreign) country. In this case, integration will give more incentive for domestic
agents to educate.
Finally, we reveal that integration may be growth damaging in the short run, while it
turns out to be growth improving in the long run. This result is illustrated by a calibra-
tion of the model for european countries. Thus, we provide more tangible evidences that
evaluate only the long-term impact of economic integration is insufficient. Our analysis
also allows to provide some policy recommandations as regards cross border externalities.
Policy makers should pursue their efforts to promote student mobility, as we emphasize
that the presence of cross border externalities is beneficial in the context of economic in-
tegration. They lead to long-term growth rate convergence across countries and may avoid
non-monotonic dynamics of the growth rates.
The paper is structured as follows. Section 3.2 presents the model. Section 3.3 deals
with autarky whereas Section 3.4 deals with integration introducing capital mobility bet-
ween the two countries. A numerical illustration is provided in Section 3.5 and finally
Section 3.6 concludes.
3.2 The model
We consider a two-country model that is an extension of Michel and Vidal (2000) in
which we introduce two production sectors : a tradable sector and a non-tradable sector. In
111
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
line with Erosa et al. (2010), the tradable sector produces a manufacturing good which can
be consumed or invested in physical capital while non-tradable sector produces services
which can be either consumed or invested in human capital. We normalize the traded good
price to unity. This two-sector production structure is a generalization of the standard two-
sector setting in which one good is a pure consumption while the other is a pure investment
good (Galor, 1992a, Venditti, 2005). We assume that investment in physical capital is
carried out only in tradable good because empirical evidences suggests that the import
component of investment is important and larger than consumption (see Burstein et al.,
2004). Since education spending is mainly supported by services in the OECD countries,
it is assumed non-tradable. 27 In this setting, the relative price of the non-traded good, PN ,
denotes the domestic real exchange rate but also the price of human capital relative to
physical capital.
The world consists of two countries which accumulate human capital and experiment
endogenous growth. In what follows, we describe the home country economy. The foreign
country economy is analogous and asterisks denote foreign country variables.
3.2.1 Production
The representative firm produces in two sectors : the tradable, and the non-tradable
one. Production in the tradable (YT ) and in the non-tradable (YN ) sector resulting from two
Cobb-Douglas production technologies, using two inputs, human capital H , and physical
capital K. Let Ki and Li, i = T,N , be respectively the quantities of capital and labor used
27. There are few papers considering multi-sector model in an international environment with human andphysical capital accumulation. Bond et al. (2003) and Hu et al. (2009) are important exceptions that studythe dynamic effect of trade. In these papers education sector is non-tradable.
112
3.2 The model
by sector i, production is given by
YT = ATKαTT H1−αT
T (3.1)
YN = ANKαNN H1−αN
N (3.2)
with αT , αN ∈ (0, 1), AT > 0 and AN > 0. To fit empirical evidence (Ito et al., 1999), we
consider :
Assumption 6. αN < αT .
Investment instantaneously transforms a unit of tradable good into a unit of installed capi-
tal and capital fully depreciates after one period. Both inputs are perfectly mobile between
the two sectors provided that :
HT +HN ≤ H, KT +KN ≤ K (3.3)
K being the total stock of physical capital and H the total amount of human capital.
Let ki = Ki/Hi be the capital intensity of sector i, hi = Hi/H be the share of human
capital allocated to sector i, i = T,N , and k = K/H the physical to human capital ratio.
Equations (3.2), (3.3) and (3.5) can be rewritten :
hT + hN ≤ 1, kThT + kNhN ≤ k (3.4)
yT = ATkαTT (3.5)
yN = ANkαNN (3.6)
113
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
where yT and yN are the production per unit of human capital in each sector.
Denotingw the wage rate,R the gross rental rate of capital and PN the price of the non-
tradable good, profit maximization over the two sectors implies that production factors are
paid their marginal product :
Rt = αTATkTαT−1t = PNtαNANkN
αN−1t (3.7)
wt = (1− αT )ATkTαTt = PNt(1− αN)ANkN
αNt (3.8)
From which we derive the physical to human capital ratios as functions of the price of the
non-tradable good :
kT t = B(PNt)1
αT−αN
kNt =αN (1−αT )αT (1−αN )
B(PNt)1
αT−αN
with B =(
αN
αT
)
αNαT−αN
(
AN
AT
) 1αT−αN
(
1−αT
1−αN
)
αN−1
αT−αN
(3.9)
And thus the input prices are :
wt = (1− αT )ATBαTPN
αTαT−αNt ≡ w(PNt)
Rt = αTATBαT−1PN
αT−1
αT−αNt ≡ R(PNt) (3.10)
3.2.2 Consumption, savings and children’s education
The economy consists, in each country, of a sequence of three life periods. In the
second period of his life, each individual gives birth to 1 + n children so that population
grows at rate n. We assume the population growth rate is the same in the two countries.
114
3.2 The model
Each generation born in period t consists of Nt identical individuals who make decisions
concerning consumption, children’s education, and savings. During childhood, individuals
make no decision : their consumption is included in their parent’s consumption. They are
reared by their parents who decide on their level of educational attainment. When adult,
they work and receive the market wage, consume, save, and rear their own children. When
old they retire, and consume the proceeds of their savings.
Individuals care about their children’s education. They exhibit a kind of paternalistic
altruism whereby they value their child’s human capital. Our modeling of intergenerational
altruism follows Glomm and Ravikumar (1992) who assume that the parental bequest
is the quality of education received by their children. The preferences of an individual
belonging to generation t are represented by :
U(ct, dt+1, ht+1) = (1− β) ln ct + β ln dt+1 + γ lnht+1 (3.11)
where ct, dt+1 and ht+1 are respectively consumption when adult, consumption when old,
and the child’s human capital ; β ∈]0, 1[ denotes individuals’ thrift and γ is the altruism
factor.
When adult, each agent born at t supplies inelastically ht+1 units of efficient labor. The
level of human capital of each adult depends on his parent’s decision on education during
his childhood :
ht+1 = bteat (3.12)
where bt is an externality, et the amount of resources a parent devotes to his child’s educa-
tion, and a ∈]0, 1[ the elasticity of the technology of human capital formation.
115
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
Let x = c, d denote individual consumption at each period of life, xN and xT be res-
pectively the spending allocated to non-traded and traded goods. Instantaneous preferences
over the two goods are defined according to :
x = xµTx1−µN (3.13)
with µ ∈ (0, 1). We denote π the consumer price index in terms of traded good. Adults
distribute their earnings that consist of labor income, wtht, among own consumption spen-
ding, investment in child’s education, and savings, st,
wtht = πtct + PNtet + st (3.14)
As expenditures in human capital et is a non-tradable good, it is assumed to be in terms
of services. Thus, PNt represents also the relative price of education services. When old,
individuals retire and consume the proceeds of their savings :
Rt+1st = πt+1dt+1 (3.15)
An individual born in period t − 1 is endowed with ht units of human capital at the be-
ginning of adulthood, and chooses et and st so as to maximize his life-cycle utility (3.11)
under his budget constraints (3.12), (3.14) and (3.15). An individual’s optimal choice is
characterized by the first order conditions :
116
3.2 The model
−1− β
πtct+
βRt+1
πt+1dt+1
= 0 (3.16)
−1− β
πtct+γa
et= 0 (3.17)
and
cTt = µπtct
PNtcNt = (1− µ)πtct
π = φ (µ) ≡ µ−µ (1− µ)−(1−µ)
(3.18)
Equation (3.16) characterizes the optimal allocation of consumption for an individual over
his lifetime. Equation (3.17) gives the optimal investment in the offspring’s human capital.
An adult reduces his consumption spending until his loss equates the increment in the
utility he derives from his child’s level of human capital out of altruism. Equations (3.18)
give the static allocation of consumption spending between the two goods.
Plugging (3.14) and (3.15) into (3.16) and (3.17) yields :
st =β
1 + γawtht (3.19)
et =γa
PNt(1 + γa)wtht (3.20)
As usual in overlapping generation models with paternalistic altruism, savings increase
117
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
with individual’s thrift and decrease with altruism. The more altruistic parents are, the
more they invest in their offspring’s education.
3.2.3 Cross-border external effects in human capital
Throughout the analysis, foreign variables are denoted by an asterisk. We assume
cross-border externalities in human capital formation. An individual’s investment in his
child’s human capital generates a positive externality for his country’s fellows. Such ex-
ternalities can be view as international spillovers in education resulting from international
student mobility. 28 For example, a visiting student can transfer his knowledge to students
in the host country and reversely, a visiting student can acquire specific learning compe-
tences when studying aboard.
We assume an externality of the form :
bt = b(pet + p∗e∗t )λet
1−a−λ and b∗t = b∗(pet + p∗e∗t )λe∗t
1−a−λ(3.21)
where b > 0, λ ∈ [0, 1 − a], p = N/(N + N∗) and p∗ = 1 − p. Since population grows
at the same rate in the two countries, p and p∗, the shares of each country in the world
population, are constant. We denote respectively et and et∗ the average levels of investment
in children’s human capital in the home and the foreign country. Since individuals are
identical within each country, in equilibrium : et = et and e∗t = et∗. The magnitude of
these cross-border external effects is given by λ. The term (pet + p∗et∗)λ is intended to
capture the strength of international spillover of knowledge. The higher λ, the more the
28. The global population of internationally mobile students more than double from 2.1 millions in 2000to 4.5 millions in 2011. According to the European commission, around 4.5 % of all European studentsreceive Erasmus grants at some stage during their higher education studies.
118
3.2 The model
home country benefits from the foreign country’s private expenditures in education.
In equilibrium, human capital depends both on domestic and foreign investment in
education and on cross-border externality in human capital formation :
ht+1 = bteta = b(pet + p∗e∗t )
λe1−λt (3.22)
Let ρt = e∗t/et be the ratio of foreign over home average investment in children’s human
capital and gt = ht+1/ht− 1 the economy growth rate. Using equations (3.20), (3.22) and
finally (3.8), we obtain :
1 + gt =γab
PNt(1+γa)AT (1− αT )kT t
αT (p+ p∗ρt−1)λ (3.23)
3.2.4 The non-tradable market clearing condition
Since there exists a non-traded good, we should consider a market clearing condition
for that good :
YNt = NtcNt +Nt−1dNt +Ntet (3.24)
This equation simply states that production equals total consumption in non-traded goods.
We can rewrite this condition with only wage, interest factor and physical to human capital
ratios :
Lemma 8. The home country non-tradable market clearing condition can be written
wt
1 + γa((1− µ)(1− β) + γa)+
Rt(1− µ)βPNt−1
(1 + n)γab(p+ p∗ρ∗t−1)λ= PNtANDkT
αN−1(kt − kT t)
(3.25)
119
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
With D = (αN (1−αT ))αN (αT (1−αN ))1−αN
αN−αT.
Proof. See Appendix 3.7.1
It can be noted that expression D is the same for both countries as we assume home and
foreign technologies have identical elasticities of substitution between production factors.
3.3 Autarky
As we first consider autarky, we rule out any interactions between countries. Invest-
ments in human capital in one country do not result in an external effect that enhances
the formation of human capital in the other (λ = 0). The human capital externality de-
pends only on the average level of education. From equation (3.21), we have with λ = 0 :
bt = bet1−a. From equation (3.22), since individuals are identical, social returns on human
capital investment are constant in equilibrium ht+1 = bet.
Young people’s savings finance the following period’s physical capital :
Kt+1 = Ht+1kt+1 = Ntst (3.26)
The labor market clears :
Ht = Ntht (3.27)
Combining (3.19), (3.20), (3.22), (3.26) and (3.27), we obtain the next period equilibrium
physical to human capital ratio :
kt+1 =β
b(1 + n)γaPNt (3.28)
120
3.3 Autarky
which depends on PNt, the price of human capital relative to physical capital in the cur-
rent period t. Using equations (3.7) and (3.8) with non-tradable market clearing condition
(3.25), we obtain :
P1
αT−αNNt =
1
B
αT
1− αT
1− αT − (1− µ)(αN − αT )αN−αT
1+γa((1− β)(1− µ) + γa) + αT
kt (3.29)
From equations (3.28) and (3.29) we finally obtain the dynamic equation characterizing
equilibrium paths :
PNt+1 =
(
β
B b(1 + n)γa
αT
1− αT
1− αT − (1− µ)(αN − αT )αN−αT
1+γa((1− β)(1− µ) + γa) + αT
PNt
)αT−αN
(3.30)
Definition 4. We define a balanced growth path (BGP) as an equilibrium where all per
capita variables grow at the same and constant rate g. This equilibrium path is such that
the relative price is constant and defined by PNt+1 = PNt = PN .
We then compute the autarkic growth rate gA on the balanced growth path.
Lemma 9. The autarkic growth factor on the balanced growth path is :
1 + gA =γab
1 + γa(1− αT )ATB
αT PA
αNαT−αN
N (3.31)
with
PAN =
[
β
b(1 + n)γa
ζ
Bη
]
αT−αN1−(αT−αN )
(3.32)
121
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
The physical to human capital ratio on the balanced growth path is :
k =
[
β
b(1 + n)γa
(
ζ
Bη
)αT−αN
] 11−(αT−αN )
≡ kA (3.33)
with ζ = 1 + (αT−αN )(1−µ)1−αT
and η = (αN−αT )αT
(1−β)(1−µ)+γa1+γa
+ 1.
Proof. See Appendix 3.7.2.
From (3.30) and (3.32), the BGP equilibrium is globally stable. 29 An increase in PNt
modifies the return of human capital (wt), which affects education spending (et) and sa-
vings (st) in the same way. Nevertheless, it also makes education spending relatively more
expensive than savings. Hence, from equation (3.28), following an increase in PNt, the
physical to human capital ratio goes up in the next time period. This means that physical
capital endowment increases relatively more than human capital endowment. From the
Rybczynski theorem, the price of the good using intensively physical capital falls. As a
result, from Assumption 6, the relative price of the non-tradable good (in terms of tra-
dable) goes up and the dynamics around the BGP is monotonous. As the growth rate is
monotonically related to the relative price of goods through equation (3.31), we can thus
claim the following :
Proposition 13. Under Assumption 6, the autarky growth rate exhibits monotonic beha-
viors.
The dynamics effect of the relative price of goods on the transitional growth rate is
little discussed in the literature. Alonso-Carrera et al. (2011) is an exception and empha-
sizes that dynamics adjustment of the relative price alters the growth rate of consumption
29. We have ∂PNt+1
∂PNt(PA
N ) = αT − αN .
122
3.3 Autarky
expenditure under particular conditions. In the infinitely lived agent model, the variation
in the relative prices generates a growth effect when there are multiple consumption goods
and non logarithmic preferences. In our model, the growth rate is endogenously determi-
ned and depends on the relative price through the returns to human capital and the price
of education spending. The growth rate dynamics is then directly driven by relative price
movements.
The following propositions contain some comparative statics results relating the long-
term relative price of goods and the growth rate to preference parameters.
Proposition 14. Under Assumption 6, the more agents value their children’s human capi-
tal the lower the relative price of non-tradable good.
Proof. We check easily from equation (3.33) that PAN is decreasing with γ.
When the non-tradable sector produces a good which can be used to invest in edu-
cation, an increase in the propensity to educate, captured by an increase in γ, has three
effects on the long-term relative price of goods PN . Two effects are directly driven by
the equilibrium on the non-tradable good market. On the one hand, a raise in γ increases
the consumption of non-tradable goods, by enhancing the education spending, such that
PN increases as well. On the other hand, a raise in γ depresses PN as it leads to a fall in
consumption spending. The third effect of γ on the relative price comes from the modifica-
tion of factors accumulation. An increase in γ favors human capital accumulation relative
to physical capital accumulation. According to the Rybzinsky theorem this increase in
human capital endowment leads to a decrease in the relative price of non-tradable good
which is produced by the human capital intensive sector. This last effect always dominates
such that a country with a higher taste for education will be characterized by a lower real
123
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
exchange rate.
We examine implications of agent’s preferences on the long-term growth rate :
Proposition 15.The more patient individuals are, the higher the growth rate. The growth
rate is first increasing and then decreasing with γ reaching a maximum in
γ =1−αT−αN+
√(1−αT−αN )2+4αN (1−αT )((1−µ)(1−β)(αN−αT )+αT )
2aαN. Moreover, under Assump-
tion 6, the growth rate is decreasing in µ.
Proof. See Appendix 3.7.3
The more altruistic individuals are, the higher their investment in children’s education
and the lower their consumption. We obtain, as in Michel and Vidal (2000), that excessive
as well as weak altruism can lead to poor growth records. The growth rate decreases with
the preference for tradable goods (µ). This is a consequence of the Stolper-Samuelson
Theorem. An increase in the propensity to consume the traded good (µ) leads to an RER
depreciation. Since the non-traded good is human capital intensive, the real depreciation
entails a fall in the wage. Then, the return to human capital decreases and so does the
growth rate.
3.4 Economic integration and growth
We consider a two-country overlapping generations world in which countries differ
in levels of patience and altruism, and in taste for non-tradable goods. We establish the
growth implications of world economic integration.
124
3.4 Economic integration and growth
3.4.1 International environment
In the integrated economy, as we assume no labor mobility between the two countries,
the labor market clearing condition of the domestic country is given as in autarky by equa-
tion (3.27). Equation (3.25) gives the non-tradable market clearing conditions for the home
country. The foreign country equations are obtained if we denote by ∗ foreign variables.
In a two-country integrated world, there are capital flows between countries and the
equality between domestic savings and domestic investment -equation (3.26)- no longer
holds. The equilibrium on the world capital market is given by :
Kt+1 +K∗
t+1 = Ntst +N∗
t s∗
t (3.34)
Dividing by the world population, the world capital market clearing condition is :
(1 + n)(
pkt+1ht+1 + p∗k∗t+1h∗
t+1
)
= pst + p∗s∗t (3.35)
With perfect capital mobility, the interest rate is the same for both countries :
Rt = R∗
t (3.36)
Using (3.10), we can determine the ratio between domestic and foreign relative prices :
P ∗Nt
PNt
=
[
AN
A∗N
] [
A∗T
AT
]
1−αN1−αT
≡ E (3.37)
This ratio reflects the bilateral real exchange rate between these two countries. We denote
ρt =e∗tet
the ratio of foreign over home average investment in children’s human capital.
125
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
The following Lemma provides a simple expression of the world capital accumulation
equation, and expressions of the physical to human capital ratios :
Lemma 10. In an integrated world, the international capital market clearing condition
is :
(pkt+1 + p∗k∗t+1ρ1−λt ) =
1
b(p+ p∗ρt)λ
(
pβPNt
γa(1 + n)+ p∗ρt
β∗P ∗Nt
γ∗a(1 + n)
)
(3.38)
and, the physical to human capital ratios are obtained from non-tradable market clearing
conditions :
kt+1 = P1
αT−αNNt+1 Bη + (1− ζ)
βPNt
b(p+ p∗ρt)λγa(1 + n)(3.39)
k∗t+1 = P ∗
Nt+1
1αT−αN B∗η∗ + (1− ζ∗)
ρλt β∗P ∗
Nt
b(p+ p∗ρt)λγ∗a(1 + n)(3.40)
The domestic price of the non-traded good is :
P1
αT−αNNt+1 =
(
pβPNtζγa
+p∗ρtβ∗P ∗
Ntζ∗
γ∗a
)
B(p+ p∗ρt)λb(1 + n)(
pη + p∗ρ1−λt η∗
) (3.41)
With ζ = 1 + (αT−αN )(1−µ)1−αT
, ζ∗ = 1 + (αT−αN )(1−µ∗)1−αT
, η = (αN−αT )αT
(1−β)(1−µ)+γa1+γa
+ 1 and
η∗ = (αN−αT )αT
(1−β∗)(1−µ∗)+γ∗a1+γ∗a
+ 1.
Proof. See Appendix 3.7.4
Introducing cross-border external effects, each country can benefit from the level of
education in the other country. Using equation (3.20) we can compute :
ρt+1 =e∗t+1
et+1
=γ∗
γ
1 + γa
1 + γ∗a
w∗t+1h
∗t+1
wt+1ht+1
1
E (3.42)
126
3.4 Economic integration and growth
Include equations (3.22) and (3.37), we have :
ρt+1 =γ∗
γ
1 + γa
1 + γ∗a
(
A∗T
AT
)
αN1−αT A∗
N
AN
ρ1−λt (3.43)
3.4.2 Steady state
Integration adds a dynamical dimension given by equation (3.43). Assuming that in-
tegration occurs at period t = 0, we consider as initial condition the state of the eco-
nomy in autarky at period −1, which gives PN0 from equation (3.41) with PN−1 = PAN ,
P ∗N−1 = P ∗A
N and ρ−1 =eA∗−1
eA∗−1
. As a result, we obtain a bi-dimensional dynamics system
that illustrates the dynamics of ρ and PN :
ρt+1 =γ∗
γ1+γa1+γ∗a
(
A∗T
AT
)
αN1−αT A∗
N
ANρ1−λt ∀t ≥ 0
PNt+1 =
pβζγa
+p∗β∗ζ∗Eρt
γ∗a
B(p+p∗ρt)λb(1+n)
pη+p∗ρ1−λt
(
A∗T
AT
) 11−αT
η∗
PNt
αT−αN
∀t ≥ 0
P1
αT−αNN0 =
(
pβPN−1ζ
γa+
p∗ρ−1β∗P∗
N−1ζ∗
γ∗a
)
B(p+p∗ρ−1)λb(1+n)(pη+p∗ρ1−λ−1 η∗)
PN−1 = PAN , P ∗
N−1 = P ∗AN and ρ−1 =
eA∗−1
eA∗−1
(3.44)
Considering that ρt+1 = ρt = ρ and PNt+1 = PNt = PN in system (3.44), we obtain :
127
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
Proposition 16. Under Assumption 6, there exists a unique non trivial stable steady state
(ρ, PN) where human capital grows at the same constant rate gw in the two economies.
ρ =
(
γ∗
γ1+γa1+γ∗a
(
A∗T
AT
)
αN1−αT A∗
N
AN
)1λ
PN =
pβζγa
+ p∗β∗ζ∗Eργ∗a
B(p+p∗ρ)λb(1+n)
pη+p∗ρ1−λ
(
A∗T
AT
) 11−αT
η∗
αT−αN1−(αT−αN )
(3.45)
and
1 + gw =γab
1 + γa(1− αT )ATB
αT PαN
αT−αNN (p+ p∗ρ)λ
In the integrated economy, the home and foreign countries grow at the same rate as
soon as there exists a positive cross border externality. This means that integration may
be growth enhancing or growth damaging in the long run depending on countries charac-
teristics. Before analyzing this question in details, next Section deals with the short-term
consequences of integration.
3.4.3 Dynamics and short-term implications
In the integrated world, the behavior of economies is driven by a two-dimensional
dynamical system. The relative price dynamics now depends on the transitional ratio ρ.
Using the dynamical system (3.44), we can analyze the local behavior of the balanced
growth path equilibrium around the steady state (PN , ρ). We have the following proper-
ties :
∂PNt+1
∂PNt(ρ, PN ) = αT−αN ;
∂PNt+1
∂ρt(ρ ; PN ) = c ;
∂ρt+1
∂PNt(ρ, PN ) = 0 ;
∂ρt+1
∂ρt(ρ, PN ) = 1−λ
128
3.4 Economic integration and growth
The eigenvalues are between 0 and 1, thus, we conclude that the converging paths around
the steady state are always monotonous.
To give intuitions about the effect of integration on relative prices and growth for the
periods following integration, we conduct a global stability analysis. We build a phase
diagram to describe the global dynamics of economy. We define the two lociEE ≡ (ρt) :
ρt+1 = ρt and PP (ρt) ≡ (ρt, PNt) : PNt+1 = PNt. From (3.44), we obtain :
ρt =
(
γ∗
γ
1 + γa
1 + γ∗a
(
A∗T
AT
)
αN1−αT A∗
N
AN
)1λ
≡ EE
And
PNt =
pζβγa
+ p∗ζ∗β∗Eρtγ∗a
B(p+ p∗ρt)λb(1 + n)
(
pη +(
A∗T
AT
) 11−αT p∗ρt1−λη∗
)
αT−αN1−(αT−αN )
≡ PP (ρt)
The EE Locus is a vertical line whereas the PP locus is a curve in plane (ρt, PNt).
Lemma 11. Under Assumption 6, if
1− a > λ >ζ∗γ∗β
ζγβ∗≡ λ
then the PP locus is first decreasing and then increasing with ρt, reaching a minimum in
ρ, denoted PminN ≡ PP (ρ). Moreover, ρ < ρ if and only if β < β. 30
Proof. See Appendix 3.7.5.
30. Values given in Appendix 3.7.5.
129
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
Then, we draw arrows representing dynamical behavior on the phase diagrams. Figure
3.2 gives the corresponding phase diagrams that depict the overall dynamic of the model
when cross border externalities are sufficiently important, i.e λ > λ 31
Non-monotonic trajectories
FIGURE 3.2 – Global Dynamics
Proposition 17. In a world with large enough cross-boarder externalities (λ ∈ (λ, 1−a)),
there exist thresholds 32 β and ρ such that integration generates a non-montonic trajectory
of the relative price if and only if one of the following set of conditions is satisfied :
— PN0 ∈ [Min(PN , PP (ρ0)),Max(PN , PP (ρ0))]
— PN0 ∈ [PminN ,Min(PN , PP (ρ0))], and
31. The condition λ > λ is not necessary to observe non-monotonic behavior. Given the non-montonicityof the PP locus, there always exist initial conditions such that non-montonic trajectories occur. This condi-tion just insures a simple representation of the phase diagram.
32. Values given in Appendix 3.7.5.
130
3.4 Economic integration and growth
— β < β and ρ0 < ρ or
— β > β, and ρ0 > ρ
Proof. See Appendix 3.7.6.
Proposition 17 identifies the situations where the converging path goes through the PP
locus. These cases illustrate the non-monotonic dynamics of the relative price and high-
light the importance to evaluate the short-term impact of economic integration. Such price
variations affect the growth rate of the economy, and thus may generate non-monotonic
dynamics in the growth rate as well. The situations where a non-monotonic dynamics in
the relative price leads to a non-monotonic dynamics of the growth rates occur when the
cross border externalities are not too large. 33
Proposition 16 and 17 have shown that integration leads to a convergence in growth rate
of integrated economies and may generate relative price and growth rate non-monotonic
trajectories. To understand why the relative price does not always evolve in a monotonous
way, we consider the relationship between relative prices and growth or relative prices and
the world ratio between physical and human capital.
We define the world physical and human capital stocks respectively byKW ≡ K+K∗
and HW ≡ H +H∗. Thus, from equations (3.22) and (3.34) we obtain :
HWt+1 ≡ b(p+ p∗ρt)
λ(Nt +N∗
t ρ1−λt )et (3.46)
KWt+1 ≡ Ntst +N∗
t s∗
t (3.47)
Perfect physical capital mobility implies that domestic and foreign interest rates converge
33. When cross border externalities are sufficiently low, the growth rate dynamics is mainly driven by thedynamics of the relative price of goods.
131
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
and then the ratio between the home and foreign relative prices is constant. This means
that when PN increases, so do education and saving, as in autarky. When PN rises, so
does P ∗N (because of perfect capital mobility), driving up global savings and the world
physical capital stock. The evolution of the world human capital stock is more complex.
Through equation (3.46), the world human capital stock does not only depend on domes-
tic education, but does also depend on the ratio between foreign and domestic education
spending (ρ). As soon as the size of externalities is low the impact of ρ on human capital
accumulation is very low compared to the one of education. Finally, a raise in PN may
lead to oscillations when this rise decreases the physical capital stock relative to the hu-
man capital stock (KW
HW ). Then, Assumption 6 implies that traded output decreases relative
to non-traded output and the relative price decreases 34. This drop in relative price drives
down education and investment and if the global capital intensity increases, traded output
rises relative to non-traded output which means that the relative price increases and so on.
Alternatively, if the global capital intensity increases following a rise in PN , the relative
price dynamics is monotonous.
The transitional dynamics exhibits striking differences compared with the case where
sectors use factor in the same proportion and where the relative price of different goods is
implicitly fixed. The model predicts that integration changes the price of education relative
to savings. Such price movements are important because they determine relative factor
accumulation between human and physical capital and also consumption choices between
tradable and non-tradable goods but also. Thus, a fall of PN induces resources to shift from
the non-tradable sector to the tradable sector.
34. Note that a modification in the endowment of capital at the world level has the same implication onproduction as in autarky (i.e. the Rybczynski theorem holds). This is because we assume that countries havethe same factor intensity between sectors and that the production of tradable goods can be aggregated atworld level as well.
132
3.4 Economic integration and growth
As in autarky, the growth rate is directly linked to the relative price of good. Never-
theless, it also depends on the ratio of foreign over home education spending, making the
analytical study of the growth rate dynamics difficult.
3.4.4 Long-term integration benefits
We examine in this section the impact of integration on the long-term growth rate. By
comparing the growth rates before and after integration in the two countries, integration is
growth enhancing when gw/gA > 1. Using equations (3.31) to (3.33) and (3.45), we can
write 1 + gw/1 + gA as a function of the foreign over domestic autarky physical to human
capital ratio, in line with Michel and Vidal (2000) :
G ≡ 1+gw
1+gA=
pη+ηp∗ρ
(
A∗T
AT
)
αT−αN1−αT
(
η∗
η
)αT−αN(
ζ∗kA∗
ζkA
)1−αT+αN
pη+
(
A∗T
AT
) 11−αT
η∗p∗ρ1−λ
αN1−αT+αN
(p+ p∗ρ)λ(1−αT )
1−αT+αN
(3.48)
G∗ ≡ 1+gw
1+gA∗ =
η∗(
A∗T
AT
) 11−αT
p
(
ATA∗T
)
αT−αN1−αT ( η
η∗ )αT−αN
(
ζkA
ζ∗kA∗
)1−αT+αN+p∗ρ
pη+
(
A∗T
AT
) 11−αT
η∗p∗ρ1−λ
αN1−αT+αN
(p+p∗ρ)λ(1−αT )
1−αT+αN
ρλ
(3.49)
We define kA∗/kA ≡ K, thus G ≡ G(K) and G∗ ≡ G∗(K). When K < 1, the fo-
reign economy is physical capital-scarce. Let us denote A1 = ( ζ∗η
ζη∗)1−αT and A2 =
(AN
A∗T
1+γ∗a1+γa
γγ∗ )
1−αTαN . The benefits from integration are then appraised in the following state-
133
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
ments :
Proposition 18. Under Assumption 6, when the home and foreign economies are charac-
terized by :A∗
T
AT∈ (Min (A1,A2) ,Max (A1,A2)), there exists a critical thresholds K,
such that, when K > K, integration is growth enhancing in the domestic country. Simi-
larly, there exists a critical thresholds K∗ such that, when K < K∗, integration is growth
enhancing in the foreign country. The two thresholds are higher than one when A1 < A2
and lower than one if A1 > A2.
Proof. See Appendix 3.7.7.
Economic integration affects growth through two channels : the cross-border exter-
nality and the relative price changes. The relative price shapes the growth rate through the
cost of education spending and the wage. In this two-sector two-factor model, under As-
sumption 6, the wage is an increasing function of the relative price of the non-traded good.
This result differs crucially from the Michel and Vidal’s one-sector setting in which the
wage increases with the capital intensity. This difference between the one-sector and the
two-sector structures drives differences between the benefits of integration. Let us consider
the case where education spending is higher in the domestic country than in the foreign
one (i.e ρ < 1 which is equivalent to the condition A∗T
AT< A2). Integration is always
growth enhancing for the foreign country as long as the domestic country is physical ca-
pital abundant (K < 1 ) and has a relatively high relative price of the non traded good in
autarky 35. In this case, integration will increase the relative price of the non-traded good in
the foreign country, so does the foreign wage and foreign education. This means that when
35. A high relative price is the brand of a strong taste for services (A1 low), or a traded sector highlyproductive (AT/AT ∗ large)
134
3.4 Economic integration and growth
A∗T
AT∈ (A1,A2) (A1 < A2), integration is growth improving for the foreign country (relati-
vely poor, with a low taste for services). A symmetric result emerges when A∗T
AT∈ (A2,A1)
(A2 < A1), with K > 1. For other intermediary cases, all scenarios may be observed : In-
tegration can be growth enhancing or reducing for the two countries, or favors one country
at the expense of the other. The different results are summarized in the following array :
ρ < 1 ⇔ A∗T
AT< A2 ρ > 1 ⇔ A∗
T
AT> A2
K > 1 K < 1 K > 1 K < 1A∗
T
AT> A1 - + gA∗ < gw < gA - + - +
A∗T
AT< A1 - + - + gA < gw < gA∗ - +
TABLE 3.1 – Long-term growth impact of integration
Economic integration can be favorable for the two countries when particular conditions
are satisfied :
Corollary 5. Under Assumptions 6, and denoting A3 = ( ηη∗)1−αt , when the economies are
such thatA∗
T
AT∈ (Min(A2,A3),Max(A2,A3)), and λ > αN/(1 − αT + αN), the two
critical thresholds satisfy K < K∗. When K ∈ (K ; K∗) integration is growth enhancing
for the two countries.
Proof. See Appendix 3.7.8.
Corollary 5 has shown that both foreign and home countries may benefit from integra-
tion in special situations. For example, let us consider that the foreign country is initially
(and definitely) more educated than the home country (ρ > 1 meaning that A∗T
AT< A2). In-
tegration improves growth in the home economy through the externality in education, if λ
is high enough. The integration will also improve growth in the foreign country if it leads
135
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
to a rise in the wage compared to the previous autarky’s situation. This rise in the foreign
wage happens only when the foreign relative price of the non-traded good increases - due
to Assumption 6. This corresponds to the case where the foreign traded productivity is low
enough (A∗T < A3AT ). To recap, both countries may benefit from integration if the home
country integrates with a foreign country highly educated but with a technological lag in
the traded sector in the case where the cross-border externality is large enough.
The long-term effects of integration do not depend on the initial condition ρ(0). Even a
ρ(0) compatible with oscillations of the growth rate in the short run may lead to a growth
improving integration in the long run. This means that after integration, the growth rate
may decrease in the short run, while the long-term global effect of integration is still
growth enhancing. Conversely, a short-term growth improvement may be compatible with
a long-term decrease in the world growth rate.
3.5 A numerical example
In this section, we derive a numerical solution for the model to illustrate the short- and
the long-term effect of economic integration. We assume that the domestic economy cor-
responds to one of the “old EU” countries (Austria, Belgium, Denmark, Finland, France,
Germany, Greece, Italy, Netherlands, Portugal, Spain, Sweden) while the foreign country
corresponds to one of the Eastern European countries that joined the EU in 2004 (Czech
Republic, Estonia, Poland). The model is calibrated using the Eurostat database, the Penn
World tables and estimations or computations provided in the literature. A summary of
calibrations and targets is provided in Table 3.5.
We assume that each period has a length of 30 years. The shares of each country in the
world population (p and p∗) is calibrated using data on the population size in 2004 (time
136
3.5 A numerical example
of integration). Using equation (3.20), we choose education preferences γa to match data
on education spendings (public and private). From equation (3.19), the discount factor β
is set to target the share of savings in GDP. Sectoral capital shares (αT , αN ) and the share
of tradable goods in consumption (µ) are set following Lombardo and Ravenna (2014). In
line with macroeconomic evidences for OECD countries, αT = 0.67 and αN = 0.33. We
use input-output table data provided by these authors to calibrate µ. These data emphasize
that countries jointed the EU in 2004 are characterized by a high preferences for tradable
good compared with most other member states. As regards sectoral TFPs, we follow Chin
(2000) by considering that average labor productivity is a proxy for sectoral TFPs. Thus,
we use data on sector labor productivity provided by Inklaar and Timmer (2012). 36 We
identify the non-tradable sector by non-market and market services and tradable sector by
manufacturing and other goods. Calibrations are in line with Hseih and Klenow (2008)
and Herrendorf and Valentinyi (2008), who emphasize that TFP gap between developed
and developing countries is not the same between sectors and developing countries are
particularly unproductive in tradable and investment goods. Finally, we abstract from po-
pulation growth fixing n = 0, and we compute the model for two different values of the
magnitude of the cross-border external effect λ : 0.05 and 0.2.
36. They use the World Input-Output Table (WIOT) of 2005 and data from the food and agriculturalorganization of the United Nations (FOAstat), and labor productivity is defined as value added per hourworked.
137
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
Parameters Target Sources
Preferences for education(γa, γ∗a)
Expenditure on publicand private educationalinstitution/GDP
EuroSTAT (average 1999-2010)
Time preferences (β, β∗) Share of savings/GDP PENN World Table(average 1999-2010)
Sectoral TFPs (AT , AN ) Sectoral labor producti-vity
Inklaar and Timmer (2012)
population (p, p∗) Population size PENN world table (2004)
Share of tradable good (µ) Share of tradable good intotal consumption
Lombardo and Ravenna (2014)
Sectoral capital shares (αT ,αN )
Capital share Lombardo and Ravenna (2014)
Countries β γa µ AT AN Pop. 2004 (million)
Austria 0,49 0,405 0,76 0,757 1,047 8,175Belgium 0,48 0,337 0,83 1,240 1,188 10,348Denmark 0,52 0,405 0,74 1,108 0,910 5,413Finland 0,50 0,322 0,49 0,966 0,824 5,215France 0,39 0,352 0,62 0,954 1,175 62,534Germany 0,42 0,336 0,71 0,998 1,033 82,487Greece 0,30 0,251 0,53 0,384 0,752 10,648Italy 0,44 0,343 0,55 0,698 0,921 58,716Netherlands 0,51 0,328 0,76 1,134 1,078 16,318Portugal 0,34 0,353 0,70 0,306 0,549 10,524Spain 0,44 0,334 0,60 0,608 0,876 43,000Sweden 0,43 0,355 0,68 0,989 1,017 8,986
Average 0,45 0,343 0,66 0,845 0,948 26,864
Czech Republic 0,44 0,273 0,77 0,256 0,557 10,246Estonia 0,40 0,309 0,79 0,197 0,499 1,342Poland 0,32 0,332 0,75 0,257 0,633 38,580
Average 0,38 0,30 0,77 0,24 0,56 16,723
TABLE 3.2 – Calibration
We compute G and G∗, given by equations (3.48) and (3.49), to determine the long-term
growth impact of economic integration for different countries. Results are summarized in
Table 3.3.
138
3.5 A numerical example
λ = 0.05
Countries Czech Republic Estonia Poland Average
G = 1+g1+gw G∗ = 1+g∗
1+gw G G* G G* G G*
Austria 0,980 2,762 0,996 3,438 0,957 2,733 0,973 2,910Belgium 0,983 3,417 0,997 4,245 0,962 3,388 0,976 3,603Denmark 0,974 3,217 0,994 4,023 0,949 3,175 0,965 3,385Finland 0,973 2,967 0,994 3,712 0,948 2,929 0,965 3,122France 0,996 3,104 0,999 3,813 0,988 3,118 0,994 3,288
Germany 0,997 2,902 1,000 3,563 0,990 2,920 0,995 3,076Greece 0,983 1,299 0,997 1,613 0,962 1,288 0,977 1,369
Italy 0,996 2,553 0,999 3,137 0,987 2,564 0,994 2,704Netherlands 0,988 3,318 0,998 4,104 0,970 3,300 0,983 3,503
Portugal 0,983 1,066 0,997 1,324 0,959 1,054 0,977 1,124Spain 0,995 2,247 0,999 2,764 0,984 2,252 0,992 2,378
Sweden 0,981 2,945 0,997 3,662 0,959 2,916 0,974 3,103Average 0,992 2,676 0,999 3,300 0,978 2,673 0,988 2,830
λ = 0.2
Countries Czech Republic Estonia Poland Average
G G* G G* G G* G G*Austria 0,922 2,598 0,985 3,399 0,840 2,398 0,895 2,677
Belgium 0,933 3,245 0,988 4,206 0,856 3,015 0,908 3,352Denmark 0,899 2,971 0,978 3,957 0,811 2,714 0,869 3,046Finland 0,897 2,735 0,977 3,649 0,809 2,497 0,866 2,804France 0,985 3,069 0,998 3,807 0,953 3,008 0,977 3,230
Germany 0,988 2,877 0,998 3,559 0,962 2,837 0,982 3,033Greece 0,939 1,241 0,988 1,599 0,880 1,178 0,915 1,283Italy 0,984 2,523 0,998 3,132 0,951 2,469 0,975 2,654
Netherlands 0,952 3,199 0,992 4,080 0,886 3,013 0,932 3,322Portugal 0,961 1,042 0,989 1,313 0,940 1,032 0,937 1,079
Spain 0,979 2,211 0,997 2,757 0,938 2,146 0,968 2,320Sweden 0,927 2,781 0,986 3,624 0,847 2,574 0,900 2,868Average 0,968 2,612 0,995 3,288 0,915 2,500 0,953 2,729
TABLE 3.3 – Long-term impact of integration on growth
Eastern European countries that joined the EU in 2004 always benefit from integration
while the long-term growth rate of other members decreases slightly. The fall in growth
139
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
in the domestic economy is low compared with the growth improvement in the foreign
country. By decomposing the domestic economy into different countries, we observe that
the gains of integration for Eastern European countries are lower when they integrate with
low-traded-TFP countries (Greece and Portugal). The reverse result is obtained when Eas-
tern European countries integrate with high-traded TFP countries (Belgium, Denmark,
Netherlands). Therefore, the TFP in the tradable sector appears as a major determinant
of the long-term impact of economic integration. Moreover, by examining more precisely
the impact of µ, we emphasize that the growth gain is higher when the consumption of
tradable good is low in the domestic country. 37 The intuition is the following : the foreign
and the domestic relative prices are positively linked in the integrated world. Therefore, a
high consumption of services in the domestic economy ensures that the relative price of
good, hence of the wage, remains sufficiently high in both countries.
The long-term growth gains are higher when λ is low, nevertheless the short-term ana-
lysis allows to emphasize that in this case convergence is observed only at very long-term.
For reasonable time horizon, the higher cross-border externalities, the higher the gains
from integration.
The numerical solution for the transitional dynamics is obtained using system (3.44).
The perfect mobility of physical capital between countries implies that the relative price
in each country jumps following integration. The figure 3.3 provides an overview of the
growth effect of economic integration for France and Czech Republic. We plot the evo-
lution of variables gIt /gA and gI∗t /g
A∗, that represent respectively the ratio of the growth
rate in the integrated economy to the autarky growth rate in France and Czech Republic. 38
37. We determine the growth effects of integration when µ changes, all other things begin equal.38. Expressions for gA and gA∗ are obtained with equation (3.31). Expressions for the growth rates in the
integrated economy are gIt = γabPNt(1+γa)AT (1 − αT )kT t
αT (p + p∗ρt−1)λ and gI∗t = γ∗ab
P∗
Nt(1+γ∗a)A
∗
T (1 −αT )k
∗
TtαT (p+ p∗ρt−1)
λρ−λt−1.
140
3.5 A numerical example
Economic integration is growth enhancing when the ratio is higher than one and growth
reducing when it is lower than one.
λ= 0.05 λ=0.2
FIGURE 3.3 – Growth impact of economic integration at t = 1
On the one hand, economic integration favors human capital accumulation in Czech
Republic because it benefits from the higher level of education spending in France. On the
other hand, with capital mobility, the two economies converge to a common world return
on physical capital. As the tradable TFP is lower in Czech Republic than in France, inte-
gration entails a fall in the relative price of education in Czech Republic which depresses
the return to human capital. When human capital externalities between countries is low
(λ = 0.05), this negative effect dominates in the short run such that economic integration
is costly for the Czech Republic. The fall in the relative price of goods is transitory. In the
long run, economic integration is beneficial for Czech Republic and the cost for France is
negligible (see Table 3.3).
When human capital externalities between countries is high (λ = 0.2), integration
immediately enhances the growth rate in Czech Republic. This is because the fall in the
relative price in this country is compensated by the positive externalities in education. The
141
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
benefit of integration is particularly important in the period following integration and tends
to reduce across time. Reversely, at the time of integration, France is negatively affected
by the lower level of education spending in Czech Republic.
For other patterns of countries, we obtain a similar result : a sufficiently high externali-
ties in education is required to observe an increase in the growth rate of the less advanced
economy at the time of integration.
3.6 Conclusion
The disagregation of the standard one-sector setting into a two-sector model with pro-
duction of traded and non-traded goods helps to account for effects of economic integra-
tion. Unlike to Michel and Vidal (2000), we identify the short-term effect of economic
integration by analyzing the behavior of non-tradable good price. We obtain that the sec-
toral traded-TFP is a crucial determinant of the growth effect of integration. From a policy
perspective, we reveal that providing funds to increase the transboundary externalities in
education is favorable : it keeps down the fall in the foreign relative price following inte-
gration and allows the poorer country to catch up faster. Moreover, it may allow to avoid
non-monotonic dynamics of the growth rate in the integrated world. We also show that
the interpretation of observations of short-term growth variations must be done with care :
an increase or decrease in the growth rate in the years following the integration does not
mean that integration will be favorable / unfavorable in long term.
142
3.7 Appendix
3.7 Appendix
3.7.1 Non-tradable market equilibrium : Proof of Lemma 8
Substituting equations (3.2) and (3.18) in equation (3.24), and dividing by Nt, we
obtain :
(1− µ)πt
(
ct +dt
1 + n
)
+ PNtet = PNtANkNαNt hthNt
Integrating the budget constraints (3.14), (3.15) and the optimal level for st and et from
equations (3.19) and (3.20) gives :
1− µ
1 + γa
(
(1− β)wtht +βRtwt−1ht−1
1 + n
)
+γa
1 + γawtht = PNtANkN
αNt hthNt (3.50)
Moreover, from the optimal choice of investment in children’s education (3.20) we know :
ht−1
ht=
1 + γa
γabwt−1
e1−at−1
bt−1
PNt−1
And thus using equation (3.22) and dividing (3.50) by ht we get :
1− µ
1 + γa
(
(1− β)wt +βRt
1 + n
(1 + γa)PNt−1
γab(p+ p∗ρt−1)λ
)
+γa
1 + γawt = PNtANk
αNNt hNt
As from equation (3.4), hN = k−kTkN−kT
, the non-tradable market clearing condition is :
1− µ
1 + γa
(
(1− β)wt +βRt
1 + n
(1 + γa)PNt−1
γab(p+ p∗ρt−1)λ
)
+γa
1 + γawt = PNtANkN
αNt
kt − kT t
kNt − kT t
143
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
From equations (3.9), we finally get the condition of the Lemma.
3.7.2 Proof of Lemma 9
Considering autarky, and thus λ = 0 and kt = kAt, the non tradable market clearing
condition (3.25) is :
1− µ
1 + γa((1− β)wt + kAtRt(1 + γa)) +
γa
1 + γawt = PNtANDkTt
αN−1 (kAt − kT t)
Substituting PNt from equations (3.9) :
1− µ
1 + γa((1−β)wt+kAtRt(1+γa))+
γa
1 + γawt = AT
αT (1− αT )
αN − αT
kTtαT−1 (kAt − kT t)
With factor prices from equations (3.7) and (3.8), we get :
1−µ1+γa
(
(1− β)(1− αT )ATkTtαT + kAt(1 + γa)αTATkTt
αT−1)
+ γa1+γa
(1− αT )ATkTtαT
= ATαT (1−αT )αN−αT
kTtαT−1 (kAt − kTt)
Dividing by kαT−1T :
(1− αT )kTt
1 + γa((1− µ)(1− β) + γa)+(1−µ)kAtαTAT = AT
αT (1− αT )
αN − αT
(kAt − kTt)
From straightforward computations, we finally obtain equation (3.29). The last task is to
compute the equilibrium growth rate gA. Using equation (3.23), we readily obtain equation
(3.31).
144
3.7 Appendix
3.7.3 Proof of Proposition 15
As PAN =
[
(
βb(1+n)γaB
)
(
αT
1−αT
1−αT−(1−µ)(αN−αT )αN−αT1+γa
((1−β)(1−µ)+γa)+αT
)]
αT−αN1−(αT−αN )
, we can define
the growth factor as a function of β, γ and µ :
1 + gA =γab
1 + γa(1− αT )ATB
αT PA
αNαT−αN
N ≡ GA(β, γ, µ)
The logarithmic derivative of GA with respect to β is :
∂lnGA(β,γ,µ)∂β
= αT
(
1β+ (1−µ)(αN−αT )
((1−µ)(1−β)+γa)(αN−αT )+αT (1+γa)
)
Which is positive if and only if
(1 + γa− µ)(αN − αT ) + αTγa
((1− µ)(1− β) + γa)(αN − αT ) + αT (1 + γa)≥ 0
The numerator is always positive. Since (1 − µ)(1 − β) < 1, the denominator is positive
for αN ≤ αT or αN ≥ αT . The growth factor is always increasing with β.
Concerning the variation of the growth rate with γ. The logarithmic derivative of GA with
respect to γ is :
∂lnGA(β,γ,µ)∂γ
= 1−αT
γ(1−αT+αN )− a(1−αT )
(1+γa)(1−αT+αN )− 1
(1−αT+αN )aαN
2
(1−µ)(1−β)(αN−αT )+αT+αNγa
= (1−αT )((1−µ)(1−β)(αN−αT )αT+αNγa)−aαN2γ(1+γa)
γ(1+γa)((1−µ)(1−β)(αN−αT )+αT+αNγa)
Which is zero for a unique positive value of γ :
γ =1−αT−αN+
√(1−αT−αN )2+4αN (1−αT )((1−µ)(1−β)(αN−αT )+αT )
2aαN.
145
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
Concerning the variations of the growth rate with µ. As
∂lnGA(β,γ,µ)∂µ
= αT (αN − αT )
(1−β)(1−αT )+αNγa+αT
(((1−µ)(1−β)+γa)(αN−αT )+αT (1+γa))(1−µαT−(1−µ)αN )
The denominator is positive from the previous analyses and the positivity of kT . Thus this
derivative is of the sign of αN − αT .
3.7.4 Proof of Lemma 10
From equations (3.19) and (3.20), we obtain st =βPNtet
γaand s∗t =
β∗P ∗Nte
∗t
γ∗a. Thus, the
world capital market clearing condition (3.35) can be written :
(1 + n)(
pkt+1ht+1 + p∗k∗t+1h∗
t+1
)
= pβPNtetγa
+ p∗β∗P ∗
Nte∗t
γ∗a
Substituting the individual level of human capital for equation (3.22) :
(1 + n)b(
pkt+1e1−λt + p∗k∗t+1e
∗1−λt
)
(pet + p∗e∗t )λ = p
βPNtetγa
+ p∗β∗P ∗
Nte∗t
γ∗a
dividing by et to write the equation as a function of ρt =e∗tet
, we obtain :
(1 + n)b(p+ p∗ρt)λ(
pkt+1 + p∗k∗t+1ρ1−λt
)
= pβPNt
γa+ p∗
ρtβ∗P ∗
Nt
γ∗a
In the integrated world, the nontraded goods market clearing condition for the home coun-
try is obtain from equation (3.25) :
kt =
wt
1+γa((1− µ)(1− β) + γa) + Rt(1−µ)βPNt−1
(1+n)γab(p+p∗ρ∗t−1)λ
PNtANDkTtαN−1 − kTt
146
3.7 Appendix
Substituting the expression of kT from equation (3.9), the factor prices wt and Rt from
equations (3.10), and simplifying by PαT−1
αT−αNNt , we obtain the expressions for kt given by
equation (3.39). The foreign ratio k∗ is deduced similarly and given by equation (3.40).
Finally, the domestic price of non-tradable goods given by equation (3.41) is obtained by
substituting equations (3.39) and (3.40) in (3.38).
3.7.5 Proof of Lemma 11
From the PP locus : limρt→0 PP (ρt) =
[
ζβγa
Bpλb(1+n)η
]
αT−αN1−(αT−αN )
≡ L0 > 0
and limρt→+∞ PP (ρt) =
ζ∗β∗
γ∗a(
A∗T
AT
) 11−αT
Bp∗λb(1+n)η∗
αT−αN1−(αT−αN )
≡ L∞ > 0. Moreover we
have :
sgn
(
∂PP
∂ρt
)
= sgn(D1(ρt))
with
D1(ρt) ≡ ε(γaβ∗ζ∗(
p∗ρtη(1− λ) + p∗ρ1−λt
(
A∗T
AT
) 11−αT η∗λ+ pη
)
−ζλβpηγ∗a− ζ(
A∗T
AT
) 11−αT η∗ γ
∗aβρλt
((1− λ)p+ p∗ρt))
We can rewrite this equation :
D1(ρt) =(
A∗T
AT
) 11−αT η∗p∗ρ1−λ
t (ζ∗γaβ∗λε− ζγ∗aβ) + ηp(ζ∗β∗γaε− λζβγ∗a)
+ εζ∗γaβ∗p∗ρtη(1− λ)− ζ η∗γ∗aβp(1−λ)
ρλt
(
A∗T
AT
) 11−αT
Since limρt→0 D1 = −∞ and limρt→+∞ D1 = +∞, the PP locus is not monotonous.
147
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
Deriving D1 with respect to ρt, we obtain :
sgn
(
∂D1
∂ρt
)
=p∗η∗
ρλt
(
A∗T
AT
) 11−αT
(ζ∗γaβ∗λ−ζγ∗aβ)+ζ∗γaβ∗p∗η+ζpη∗γ∗aβ
(
A∗T
AT
) 11−αT λ
ρ1+λt
Under the following sufficient condition ζ∗γaβ∗λ > ζγ∗aβ, D1 is strictly increasing in ρt.
Consequently, PP is decreasing and then increasing in ρt and achieves a minimum in ρ.
The threshold ρ is lower than ρ when D1(ρ) > 0. As D1(ρ) is decreasing in β and positive
when β = 0, there exists a β under which ρ < ρ.
3.7.6 Proof of Proposition 17
As integration occurs in period 0, education decisions made in autarky at period −1 are
considered as initial conditions (ρ−1) for the integrated economy, as the autarky relative
prices PN−1 and P ∗N−1. According to the resulting (ρ0, PN0) different trajectories may
emerge. More precisely, non-monotonic behaviors occur when the trajectory goes through
the PP locus where ∆PN = 0. In this case the trajectory has a zero slope.
The proof proceeds by reduction to the absurd. When PP (ρ0) < PN0 < PN , the
economy starts in a regime where the relative price goes down. Thus, the price at the
second period, PN1, is characterized by PN1 < PN0 < PN . The only possible way to
converge to PN is that the economy cuts the isoline PP to achieve the area where the
price increases. Reversely, when PN < PN0 < PP (ρ0), the economy starts in a regime
where the relative price goes up. The price at the second period is characterized by PN1 >
PN0 > PN . The only possible way to converge to the long-term equilibrium is that the
economy achieves the area where the price falls.
148
3.7 Appendix
We describe the relative price dynamics in all the phase diagram regions in the case
where 1− a > λ > ζγ∗aβζ∗γaβ∗ . As previously, the proof proceeds by reduction to the absurd.
— When PN0 < PminN , the economy starts in a regime where the relative price goes up
and stays until the convergence.
— When PN0 ∈ (PminN , PN) :
• For β < β :
i) PN0 > PP (ρ0) the economy starts in a regime where the relative price
goes down. Thus, PN1 < PN and the only possible way to converge is
that the economy cuts the isoline PP to achieve the area where the price
increases.
ii) PN0 < PP (ρ0) and ρ0 > ρ the economy starts in a regime where relative
price and ratio ρ go up and the locus PP is increasing in ρ. Thus, there is
no dynamics changes.
iii) PN0 < PP (ρ0) and ρ0 < ρ the economy starts in a regime where the
relative price and ratio ρ go up. Nevertheless, in this area the locus PP is
decreasing in ρ. Thus, the economy cuts the isoline and goes in the area
where the relative price goes down. If the increase during the first periods
is sufficient, i.e PNt > PN when the economy achieves the decreasing
area, the economy converges without experiences other regime changes.
Reversely, if PNt < PN the only possible way to converge is that the
economy cuts again the isoline PP to achieve the area where the price
increases. In this particular case the economy converges in an oscillatory
way.
• For β > β, we obtain a symmetric result :
149
3 Short-and long-term growth effects of integration in two-sector economies with non-tradablegoods
i) PN0 > PP (ρ0) non-monotonous path.
ii) PN0 < PP (ρ0) and ρ0 < ρ monotonous path.
iii) PN0 < PP (ρ0) and ρ0 > ρ non-monotonous path with possible oscilla-
tions.
— When PN0 > PN :
i) PN0 > PP (ρ0) the economy starts in a regime where the relative price
goes down. As the economy starts under the PP locus it converges monoto-
nously.
ii) PN0 < PP (ρ0) the economy starts in a regime where the relative price
goes up, the only possible way to converge is that the economy cuts the isoline
PP to achieve the area where the price goes down.
3.7.7 Proof of Proposition 18
Using equations (3.48) and (3.49) we have the following properties : G(K) is a increa-
sing function of K with G(0) > 0 and G(∞) = ∞ and G∗(K) is a decreasing function of
K with G∗(0) = ∞ and G∗(∞) > 0. Moreover, we have :
— G(1) < 1 and G∗(1) > 1 when ρ < 1 and ζη∗(
A∗T
AT
) 11−αT > ζ∗η. In this case there
exists K > 1 over which integration favors the domestic country and K∗ > 1 under
which integration favors the foreign country.
— G(1) > 1 and G∗(1) < 1 when ρ > 1 and ζη∗(
A∗T
AT
) 11−αT < ζ∗η. In this case there
exists K < 1 over which integration favors the domestic countries and K∗ < 1 under
which integration favors the foreign country.
150
3.7 Appendix
Using (3.33) and (3.45), ρ < 1 and ζη∗(
A∗T
AT
) 11−αT > ζ∗η can be compatible with K > 1.
Similarly, ρ > 1 and ζη∗(
A∗T
AT
) 11−αT < ζ∗η can be compatible with K < 1.
3.7.8 Proof of Corollary 5
From equations (3.48) and (3.49), there exists K which satisfies G(K) = G∗(K) with :
K =
(
A∗T
AT
) 11−αT ζη∗ρ
−λαN
ζ∗η
At this particular point, the two countries are characterized by the same autarky growth
rate and the ratio between the autarky and the integrated growth rate is given by :
G(K) =
pη +(
A∗T
AT
) 11−αT η∗p∗ρ
1−λ(1−αT+αN )
αN
pη +(
A∗T
AT
) 11−αT η∗p∗ρ1−λ
αN1−αT+αN
(p+ p∗ρ)λ(1−αT )
1−αT+αN ≡ M
Integration is growth enhancing for the two countries if M > 1. We examine how ¯ρ affects
the function M.
When ρ = 1, we have M = 1. Take the log of the function and derive according to ρ, we
obtain :
sgn(
∂ lnM
∂ρ
)
= −(1− λ)αN
(
(
A∗T
AT
) 11−αT η∗ρ−λ − η
)(
pη + p∗(
A∗T
AT
) 11−αT η∗ρ
1−λ(1−αT+αN )
αN
)
+ (αN − λ(1− αT + αN ))
(
pη + p∗(
A∗T
AT
) 11−αT η∗ρ1−λ
)(
(
A∗T
AT
) 11−αT η∗ρ
−λ(1−αT+αN )
αN − η
)
Focus on the case where cross-border externalities in human capital are sufficiently high,
λ > αN/(1− αT + αN). When the domestic country is patient compared with the foreign
(η∗ > η) and that the integrated economy converges to ρ < 1, M is a decreasing function
of ρ. Consequently, M > 1 ∀ ρ ∈ (0, 1) and there exist two critical thresholds K and
K∗, such that when K ∈ (K ; K∗), integration is growth enhancing for the two countries.
151
4 Environmental policy and growth in a model with endogenous environmental awareness
Similarly, integration may be growth enhancing for the two countries when ρ > 1 and
η∗ < η. In this case M, is an increasing function of ρ such that M > 1 ∀ ρ ∈ (1,+∞).
152
Chapter 4
Environmental policy and growth in a model with
endogenous environmental awareness
4.1 Introduction
The link between growth and the environment is a fundamental issue in environmen-
tal economics, as highlighted in the literature reviews of Brock and Taylor (2005) and
Xepapadeas (2005). One of the main questions raised is the role of environmental po-
licy in attaining a sustainable development, where economic growth is compatible with a
non-damaging environment. To achieve such a goal, policy makers have several economic
levers. The most obvious instruments are pollution taxation and public pollution abatement
(e.g. water treatment, waste management, investment in renewable energy, conservation of
forests...), introduced to reduce environmentally harmful activities.
But the governments may also invest in other types of policy tools that aim to modify
households’ behaviors. In this regard, education, by raising environmental conscious, can
be used as an indirect intervention in favor of environment. This idea is supported by inter-
national organizations. For example, OECD (2008) refers to education as “one of the most
powerful tools for providing individuals with the appropriate skills and competencies to
become sustainable consumers” , while the United Nations declares the decade 2005-2014
as the “UN Decade of Education for Sustainable Development”. 39 Recently, OECD (2007,
39. See resolution 57/254 of United Nations General Assembly of 2002.
153
4 Environmental policy and growth in a model with endogenous environmental awareness
2010) also suggest to combine several levers of environmental policy. Given the “multi-
aspect” nature of environmental issues, instruments are likely to mutually reinforce each
other. In particular, they emphasize that pollution taxation may not have the expected re-
turns without complementary actions focusing on households’ behaviors.
In the light of the development of such policies directed to consumers’ behaviors, it
seems particularly relevant to consider the role of agents’ green preferences, which deter-
mine how they respond to pollution. Moreover, recent studies as European Commission
(2008) point out that households are becoming more aware of environmental issues and of
their role in environmental protection since the recent decades, reflecting that these pre-
ferences evolve over time. The purpose of this paper is thus to study how environmental
policy affects growth performance, when individual preferences for the environment are
endogenous.
A number of studies deals with the link between environmental policy and growth,
however no consensus exists. Ono (2003a) underlines the intergenerational effect of en-
vironmental taxation. In his paper, tax reduces the polluting production, but improves the
level of environmental quality bequeathed to future generation, such that there exists an
intermediary level of tax that enhances the long-term growth. Other contributions examine
this issue by considering that human capital accumulation is the engine of growth. In a Lu-
cas model, Gradus and Smulders (1993) emphasize that an improvement in environmental
quality has a positive effect on long-term growth when pollution affects directly human
capital accumulation. More recent papers underline that a tighter environmental tax can
favor education and hence growth at the expense of polluting activities, even without such
assumption. For example, in a model with a R&D sector reducing pollution, Grimaud and
Tournemaine (2007) obtain this result as the tax increases the relative price of the polluting
154
4.1 Introduction
good. For finite lifetime, Pautrel (2011) finds that an increase in tax enhances growth as
long as the abatement sector is more human capital intensive that the final output sector,
while in Pautel (2012) this result arises because pollution stems from physical capital. 40
We depart from these papers in two major ways. First, we analyze an environmental
policy with possible “instrument mixes”. The government can implement a tax on pollu-
tion and recycle tax revenues in two types of environmental support : a direct one through
public pollution abatement and a more indirect one through education subsidy.
Second, we assume that agents’ preferences for the environment are endogenous. More
precisely, we consider that both the individual human capital and the level of pollution af-
fect green preferences positively, as supported by literature. A wide range of empirical
studies identifies education as a relevant individual determinant of environmental prefe-
rences (see Blomquist and Whitehead, 1998, Witzke and Urfei, 2001, European Commis-
sion, 2008). The intuition is that the more educated agent is, the more she is informed
about environmental issues, and the more she can be concerned about environmental pro-
tection. Likewise, environmental issues, as climate change or air pollution, harm welfare
and push households to realize they are facing a serious problem and hence the need to
react. Among other channels, pollution affects agents’ well-being by damaging their health
status (through mortality and morbidity) and by depreciating the environmental quality be-
queathed to their children. Accordingly Schumacher (2009) highlights that when pollution
is high, agents are more likely to be environmentally concerned and to act for the environ-
ment.
Considering an endogenous environmental awareness, our analysis is also related to
the recent contribution of Prieur and Bréchet (2013), in which green preferences depend
40. Pautel (2012) has a similar result than Gradus and Smulders (1993), in which pollution is due tophysical capital, but it does not require that pollution affects directly education.
155
4 Environmental policy and growth in a model with endogenous environmental awareness
on human capital. The authors emphasize that the economy may be caught in a steady state
without economic growth, while education policy can be used to achieve an asymptotic ba-
lanced growth path with sustainable growth. Here, we extend this paper considering that
education choices are not exogenous but stem from paternalistic altruism and that environ-
mental preferences are driven by both human capital and pollution. 41
In our overlapping generations model, growth is driven by human capital accumula-
tion and environmental quality. Human capital depends on education spending chosen by
altruistic parents, while the law of motion of the environment is in line with John and Pec-
chenino (1994). Production creates pollution flow, which damages environmental quality,
whereas abatement activities improve it. To well identify consumer’s environmental pre-
ferences, we use an impure altruism à la Andreoni (1990), where the contribution to the
public good arises from private preferences for this good (pure altruism) and from a joy
of giving. With this formalization, public and private contributions are not perfect substi-
tutes, as supported empirically, and the analysis of policy is meaningful. 42 Consequently,
we consider two different incentives to explain pollution abatement : the level of environ-
mental quality and the contribution itself to the environment.
With the present model, two regimes are distinguished : a regime with private contribu-
tion to pollution abatement and a regime without private contribution to this good. Under
reasonable conditions, the economy converges to a sustainable balanced growth path (he-
reafter BGP), where both environmental quality and human capital grow. Depending on
the share of public spending in public maintenance and the level of the tax, the BGP can
41. For a paper considering the effect of environmental quality on green concerns, see Schumacher andZou (2009). Nevertheless, they assume that environmental quality has a discrete impact on preferences.
42. In other words, when the government increases spending for the public good, there is not a completecrowding out effect on private contribution. Such impure altruism and imperfect crowding-out are supportedby empirical evidence. See Ribar and Wilhelm (2002) or Crumpler and Grossman (2008) for a review oncharitable giving and Menges et al. (2005), on environmental contribution.
156
4.1 Introduction
be with or without private maintenance. We reveal that when the BGP is characterized by
private pollution abatement, endogenous environmental concerns may generate damped
oscillations. This occurs when environmental preferences are highly sensitive to econo-
mic changes. Specifically, the feedback effect of human capital and environmental quality
on green preferences influences the trade-off between private choices, which generates
oscillations. Such complex dynamics leads to significant variations in the welfare across
generations, which correspond to intergenerational inequalities along the convergence path
(see Seegmuller and Verchère, 2004).
While the effect of environmental policy is generally studied in the long run, we under-
line that the short-term analysis represents also a crucial issue. Thus, we study the effect
of the policy in both the short-and-long run, and especially if it can avoid intergenerational
inequalities and favor the long-term growth rate. We do not focus on the welfare analysis,
which is not analytically tractable in our setting but we give intuitions about the economic
impact of environmental policy instrument.
We emphasize that an increase in tax can enhance growth and remove inequalities as
long as the tax revenue is well allocated. More precisely, an intermediary allocation of the
budget between public maintenance and education ensures that the economy converges
to a BGP without private maintenance and with a sufficiently high support to education.
Indeed, in this regime, environmental maintenance is entirely supported by public autho-
rities. Consequently, there is no trade-off between private choices and oscillations never
occur. To achieve this regime, and hence avoid short-term issue, the share of budget in
public maintenance has to be high enough. However, we show that an intermediary allo-
cation of tax revenue is required to reach the highest long-term growth rate of both human
capital and environmental quality. In this way, the budget devoted to pollution abatement
157
4 Environmental policy and growth in a model with endogenous environmental awareness
is sufficiently high to be in the regime without private maintenance, where environmental
quality is good, but education support is also sufficient so that the negative effect of tax on
available income is more than offset by the positive effect of education subsidy. Thus, our
study confirms that the implementation of policies mix is crucial for environmental issues.
Our study confirms that the implementation of policies mix is crucial for environmental
issues.
The paper is organized as follows. In Section 4.2, we set up the theoretical model.
Section 4.3 focus on the BGP and the transitional dynamics. In Section 4.4, we examine
short-and long-term implications of environmental policy. Finally, Section 4.5 concludes.
Technical details are relegated to an Appendix.
4.2 The model
Consider an overlapping generations economy, with discrete time indexed by t =
0, 1, 2, ...,∞. Households live for two periods, childhood and adulthood, but take all de-
cisions during their second period of life. At each date t, a new generation of N identical
agents is born (N > 1). We assume no population growth.
4.2.1 Consumer’s behavior
Individual born in t − 1 cares about her adult consumption level ct, her child’s hu-
man capital ht+1, the current level of environmental quality Qt and the future environ-
ment through altruism. To properly represent this latter environmental preference (see for
example Menges et al., 2005), we use an impure altruism in line with Andreoni (1990).
Agent gets welfare for the total level of public good (i.e. the future environmental quality
Qt+1) but also from her action to contribute to this good (i.e. environmental maintenance
158
4.2 The model
mt). With the joy of giving for mt, public and private contributions are no longer per-
fect substitutes, such that the controversial neutrality result of policy predicted by purely
altruistic models, does not hold. Preferences are represented by the following utility func-
tion of a representative agent :
U(ct,mt, ht+1, Qt+1) = ln ct + γ1t ln(ε1 mt + ε2 Qt+1) + γ2 lnht+1 + γ3 lnQt (4.1)
with γ1t, γ2, γ3, ε1 and ε2 > 0.
The parameter γ3 captures taste for the current environmental quality. It corresponds
to a usual well-being due to the environment but is taken as given by the agent and not
related with altruism concerns. On the other hand, the factor γ2 represents the preference
for her child’s human capital. As a private good, this choice is only determined by paterna-
listic altruism such that parents finance the child’s education, as in Glomm and Ravikumar
(1992).
The weight γ1t captures the environmental awareness. We consider that these envi-
ronmental preferences are affected negatively by the level of environmental quality (Qt)
and positively by the individual human capital (ht), as supported by literature. Pollution,
affecting welfare, has an impact on environmental behaviors : when pollution is high,
agents are more likely to be concerned by the environment and to act in favor of it, as
underlines Schumacher (2009). The worst environmental quality, the more the individual
is able to realize the badness of the situation and therefore the more she has an incentive
to protect the environment. At the same time, empirical behavioral economics literature
identifies education as a determinant of the contribution to the environment ((see Blom-
quist and Whitehead, 1998, Witzke and Urfei, 2001). The economic intuition is that the
more an agent is educated, the more she may be informed about environmental issues
159
4 Environmental policy and growth in a model with endogenous environmental awareness
and their consequences, and thus the more she can be concerned about it. We assume that
γ1t = γ1(ht, Qt) where γ1 is increasing and concave with respect to h, and decreasing and
convex with respect to Q. In particular, we consider the following functional form : 43
γ1t ≡βht + ηQt
ht +Qt
(4.2)
with parameters β, η ∈ [0, 1] and β > η. The parameters β and η embody respectively
the weight of human capital and of environmental quality in green preferences. Let us
underline that when β = η, the environmental awareness is constant.
During childhood, individual does not make decisions. She is reared by her parents and
benefits from education. When adult, she supplies inelastically one unit of labor remune-
rated at the wage wt according to her human capital level ht. She allocates this income
to consumption ct, education per child et and environmental maintenance mt. 44 Further-
more, the government can subsidy education at the rate 0 6 θet < 1, reducing the private
cost of education. The budget constraint for an adult with human capital ht is :
ct +mt + et(1− θet ) = wtht (4.3)
The human capital of the child ht+1 is produced with the private education expenditure
et and the human capital of the parents ht :
ht+1 = ǫ etµht
1−µ (4.4)
43. For a similar form, see Blackburn and Cipriani (2002) who use it to model the effect of pollution onlongevity.
44. See Kotchen and Moore (2008) for empirical evidences of private provision of environmental publicgoods.
160
4.2 The model
with ǫ > 0, the efficiency of human capital accumulation. The parameter 0 < µ < 1 is
compatible with endogenous growth and captures the elasticity of human capital to private
education, while 1 − µ represents the share of human capital resulting from intergenera-
tional transmission within the family.
The law of motion of environmental quality is defined by :
Qt+1 = (1− α)Qt + b(mt +Mt +NGmt )− aYt (4.5)
where α > 0 is the natural degradation of the environment and Yt represents the pollution
flow due to production in the previous period. The parameter a > 0 corresponds to the
emission rate of pollution, while b > 0 is the efficiency of environmental maintenance.
The abatement activities are represented by a Cournot-Nash equilibrium approach. Each
agent determines her own environmental maintenance (mt), taking the others’ contribution
(Mt) as given. The government can provide public environmental maintenance NGmt > 0,
which has the same efficiency than the private one. Following the seminal contribution
of John and Pecchenino (1994), Q is an environmental quality index with an autonomous
value of 0 in the absence of human intervention. This index may embody, for example, the
inverse of the concentration of greenhouse gases in the atmosphere (like the chlorofluoro-
carbons, CFCs), or a local environmental public good such as the quality of groundwater
in a specific area. 45
The consumer program is summarized by :
maxet,mt
, U(ct,mt, ht+1, Qt+1, Qt) = ln ct+γ1t ln(ε1 mt+ε2 Qt+1)+γ2 lnht+1+γ3 lnQt
45. For the analysis, we consider Q > 0. This assumption is standard in the literature, see Ono (2003b)or Mariani et al. (2010).
161
4 Environmental policy and growth in a model with endogenous environmental awareness
(4.6)
s.t ct +mt + et(1− θet ) = wtht
ht+1 = ǫ etµht
1−µ
Qt+1 = (1− α)Qt + b(mt +Mt +NGmt )− aYt
with mt > 0.
4.2.2 Production
Production of the consumption good is carried out by a single representative firm. The
output is produced according to a constant returns to scale technology :
Yt = AHt (4.7)
where Ht is the aggregate stock of human capital and A > 0 measures the technology
level. Defining yt ≡ Yt
Nas the output per worker and ht ≡ Ht
Nas the human capital per
worker, we have the following production function per capita : yt = Aht.
The government collects revenues through a tax rate 0 6 τ < 1 on production, which is
the source of pollution. By assuming perfect competition, the profit-maximization problem
yields the following factor price :
wt = A(1− τ) (4.8)
In our model, pollution is a by-product of the production process, which uses human
capital only for the sake of simplicity. We assume that A contains the intensive-polluting
162
4.2 The model
factors (e.g. physical capital) that are considered to be exogenous here. Thus, we can
interpret the technology level A as an index of pollution intensity. 46
4.2.3 The government
The design of environmental policy represents a major challenge for governments.
Among other reports, OECD (2007 and 2008) recommends the revenue recycling of tax
on polluting activities in order to complete the governmental action. This kind of policy
is observable in several countries. For example, in France, the government implements
a general tax on polluting activities (TGAP) and transfers revenues to the French En-
vironment and Energy Management Agency (ADEME) that funds activities in favor of
environment. While environmental policy is often studied through taxation in theoretical
literature, there exist several policy levers. Policy makers can support direct environmen-
tal actions (e.g. conservation of forests and soils, water treatment, waste management), but
also more indirect actions attempting to change behaviors.
In this model, in order to study such environmental policy, we consider the following
policy scheme : since pollution is a by-product of the production process, the government
taxes the output at rate τ and the public budget is spent on public environmental mainte-
nance NGmt or/and on education subsidy θet . The direct impact of an education policy is
to improve human capital accumulation and hence to raise income. Nevertheless, it also
generates an indirect impact on green preferences as human capital affects positively en-
vironmental awareness. In this sense, support in education can be viewed as an indirect
46. We thank an anonymous referee for suggesting this interpretation. Note that introducing physicalcapital accumulation would make the analysis much more complex, as it introduces a new dimension in thedynamics.
163
4 Environmental policy and growth in a model with endogenous environmental awareness
environmental policy. The government’s budget is balanced at each period, such that :
N(θet et +Gmt ) = τYt (4.9)
We define the share of public expenditure devoted to public maintenance 0 6 σ 6 1, and
to education subsidy (1− σ), assumed constant :
σ =NGm
t
τYt; 1− σ =
Nθet etτYt
(4.10)
Thus, fiscal policy is summarized by two instruments τ ; σ, taken as given by consu-
mers. 47
4.2.4 Equilibrium
The maximization of the consumer program (4.6) leads to the optimal choices in terms
of education and maintenance in two regimes : an interior solution, where individuals
invest in environmental protection mt > 0 (hereafter pm) and a corner solution without
private contribution to the environment mt = 0 (hereafter npm). The Nash intertemporal
equilibria are given by : 48
mt =
γ1tc1Aht(1−τ)−ε2(1+γ2µ)[(1−α)Qt+ANht(bστ−a)]γ1tc1+c2+ε2bN(1+γ2µ)
pm
0 npm
(4.11)
47. We may also consider a policy scheme where Gmt is the only endogenous variable and the policyinstruments are τ and θ. In this case, we have τYt = Nθet +NGmt. and education policy does not dependon pollution tax. This formalization does not change significantly the analysis and the results.
48. See details in Appendix 4.6.1.
164
4.2 The model
et =
γ2µ[c3A(1−τ)ht+ε2((1−α)Qt+ANht(bστ−a))]γ1tc1+c2+ε2bN(1+γ2µ)
+ (1− σ)τAht pm
Aht[γ2µ(1−τ)+τ(1−σ)(1+γ2µ)]1+γ2µ
npm
(4.12)
where c1, c2 and c3, three positive constants defined by c1 ≡ ε1 + ε2b ; c2 ≡ ε1(1 + γ2µ)
and
c3 ≡ ε1 + ε2bN .
Education spending depends positively on environmental quality. The better the envi-
ronment, the lower the optimal amount of maintenance activities, as a result, individual
can devote more resources to educate her child.
The public policy instruments shape education and abatement spendings differently.
An increase in tax implies a negative income effect (wage decreases) but still favors edu-
cation spending when public expenditure is sufficiently devoted to education subsidies (σ
low). 49 Conversely, an increase in tax always affects negatively maintenance activities.
In addition to the negative income effect, the tax increases the public pollution abatement,
which crowds out private maintenance. Nevertheless, public spending substitutes only par-
tially to the private one due to the direct benefit from contribution to the environment.
Remark 1 Without a joy-of-giving motive for the environment (i.e. ε1 = 0) :
— If all the public expenditures are devoted to pollution abatement (σ = 1), there is a
perfect crowding-out of private maintenance such that environmental policy has no
effect on the environment.
— If the budget is divided between the two types of expenditure (σ < 1), the fall in
49. ∂et∂τ
> 0(
resp. ∂et∂τ
< 0)
, when σ < γ1tc1+c3γ1tc1+c3+ε1γ2µ
(
resp. 1 > σ > γ1tc1+c3γ1tc1+c3+ε1γ2µ
)
.
165
4 Environmental policy and growth in a model with endogenous environmental awareness
private maintenance outweighs the increase in public maintenance. Consequently,
the overall maintenance decreases with the environmental policy.
Such cases are in contradiction with the empirical and experimental literature which only
recognizes a partial crowding-out of private contribution by government expenditures. 50
Therefore, the joy-of-giving motive (ε1) is necessary for a meaningful policy analysis.
Studying endogenous growth, we introduce a green development index Xt, equal to
environmental quality per unit of human capital : Xt ≡ Qt
ht. Thus, the environmental awa-
reness, given by (4.2), can be rewritten :
γ1t =β + ηXt
1 +Xt
(4.13)
Using equation (4.11), we deduce the condition such that the regime without private
environmental maintenance occurs :
Xt >A [γ1tc1(1− τ)− ε2N(1 + γ2µ)(bστ − a)]
ε2(1 + γ2µ)(1− α)(4.14)
When this inequality is satisfied, the level of environmental quality is so high and/or the
level of human capital so low, that the private abatement of pollution is given up. Policy
favors the occurrence of this regime. When τ is positive, the economy is more likely to be
characterized by no maintenance activities, especially if σ is positive, since public mainte-
nance partially substitutes for private one. But when revenue recycling is entirely devoted
to education (σ = 0), agent still replaces some abatement activities by education, as the
last becomes relatively less costly. From equation (4.14), we derive that households invest
50. See e.g. Ribar and Wilhelm (2002), Menges et al. (2005) or Crumpler and Grossman (2008).
166
4.3 Balanced growth path and transitional dynamics
in environmental protection when Xt ∈ (0,Λ), with Λ > 0. 51
Using the human capital accumulation (4.4), the environmental quality process (4.5),
and the first order conditions (4.11) and (4.12), we finally obtain the dynamic equation
characterizing equilibrium paths :
Definition 5. Given the initial condition X0 = h0
Q0> 0, the intertemporal equilibrium is
the sequence (Xt)t∈N which satisfies, at each t, Xt+1 = F(Xt), with :
F(Xt) =
(1−α) Xt [γ1t c1+c2]+AN [γ1t c1b(1−τ)+(γ1c1+c2)(bτσ−a)]
ǫ[γ2µA c3(1−τ)+γ2µε2[(1−α)Xt+AN(bστ−a)]+(1−σ)τA(γ1t c1+(1+γ2µ)c3)]µ[γ1t c1+(1+γ2µ)c3]
1−µ pm
(1−α)Xt+AN(bστ−a)
ǫ[
γ2µA(1−τ)+(1−σ)τA(1+γ2µ)1+γ2µ
]µ npm
(4.15)
4.3 Balanced growth path and transitional dynamics
We examine in this section the existence of a BGP equilibrium characterized as :
Definition 6. A balanced growth path (BGP) satisfies Definition 1 and has the following
additional properties : the stock of human capital and environmental quality grow at the
same and constant rate gi, with subscripts i = pm, npm denoting respectively the re-
gime with Private Maintenance and the regime where there is No Private Maintenance.
This equilibrium path is such that the green development index Xt is constant and defined
by Xt+1 = Xt = Xi.
From Definitions 5 and 6 and equations (4.13) and (4.14), we emphasize the properties
of the dynamic equation F and deduce the existence of the BGP Xi corresponding to the
51. See technical details in Appendix 4.6.1.
167
4 Environmental policy and growth in a model with endogenous environmental awareness
solutions of equation Xi = F(Xi) :
Proposition 19. When β > η and σMin(τ) < σ < σMax(τ) for all τ ∈ [0, 1), there exists
a unique positive BGP (Xi), such that, according to a critical threshold σ(τ) :
• When σ > σ(τ), the BGP is in the regime without private maintenance (npm).
• When σ < σ(τ), the BGP is in the regime with private maintenance (pm).
where σ(τ) is a decreasing function of τ , with limτ→0 σ(τ) = +∞ and limτ→1 σ(τ) =
a/b.
Proof. See Appendix 4.6.2.
The balanced growth path corresponds to a sustainable development, where both hu-
man capital and environmental quality improve across generations. The policy makes pos-
sible that such a sustainable BGP exists without private abatement. If the share of public
spending devoted to environmental protection (σ) is sufficiently high, households may stop
investing in private maintenance in long run, as underlined in Proposition 19. However,
when environmental awareness (γ1) is too high, the tax rate (τ ) required to the existence
of the regime without private maintenance is important.
For the rest of the paper, we set :
Assumption 7. For all τ ∈ [0, 1), we assume that σMin(τ) < σ < σMax(τ).
This assumption guarantees the existence of a sustainable BGP. The inequality implies
that human capital accumulation and environmental maintenance are sufficiently efficient.
More precisely, it entails that the efficiency of maintenance activities, devoted to the pro-
tection of the environment, is higher than the weight of pollution flow in environmental
quality, which is only a side-effect of the production process (b > a). The assumption
168
4.3 Balanced growth path and transitional dynamics
restrains some policy schemes : for low level of tax, all allocation σ are possible, while for
high level, extreme allocations between public spendings are excluded.
From the study of the dynamic equation, we derive the stability properties of the BGP
presented in Proposition 20. When the BGP is in the regime without private maintenance,
we obtain an explicit solution whose dynamics is easily deduced. However, to analyze the
stability of the equilibrium in the regime with private maintenance, we normalized Xpm to
one, using the scaling parameter ǫ.
Proposition 20. Under Assumption 7 and β > η :
— The BGP in the regime without private maintenance, Xnpm, is globally and monoto-
nously stable.
— The BGP in the regime with private maintenance, Xpm, is locally stable and for
N > N , there exists a β ∈ (0, 1] such that :
— when β < β, the convergence is monotonous.
— when β > β, the convergence is oscillatory.
Proof. See Appendix 4.6.3.
Figure 4.4 provides an illustration of the cases identified in Proposition 20.
As underlined in Proposition 20, the economy may display damped oscillations. be-
cause of endogenous concerns. 52 The emergence of complex dynamics is explained by the
feedback effect between green development index and environmental awareness, which af-
fects the trade-off between education and maintenance.
In the absence of private maintenance, this trade-off does not exist as households focus
on education and the dynamics is always monotonous. Whereas when agent invests in en-
52. At the limit case β = η, γ1 is exogenous and the dynamics is always monotonous. The proof isavailable upon request.
169
4 Environmental policy and growth in a model with endogenous environmental awareness
Xt+1
Xt0
Λ
FXnpm
Xt+1 = Xt
when σ > σ(τ)
Xt+1
Xt0
β > β
F ′
X′
pm
Λ′
β < β
Λ′′
F ′′
X ′′pm
Xt+1 = Xt
when σ < σ(τ)
FIGURE 4.4 – Dynamics when N > N
vironmental protection, cyclical convergence may occur and can be described as follows.
An increase in environmental awareness γ1t encourages private maintenance investment at
the expense of education spending. For the next generation, it generates a fall in human ca-
pital ht+1, a raise in environmental quality Qt+1 and hence a decrease in green preferences
γ1t+1. These modifications entail multiple effects on the private choices. They all shape
negatively private maintenance mt+1 whereas the impact is ambiguous for education et+1
. Indeed, education spending is affected positively by the fact that private maintenance be-
comes less needed (from the raise in Qt+1) and less wanted (from the decrease in γ1t+1),
while it is affected negatively by an income effect due to the fall in human capital ht+1.
Actually, the economy displays oscillations as long as the positive impacts on education
exceed the negative income effect. Note that the opposite variation of human capital ht+1
acts as a brake on oscillations, i.e. a stabilizing effect, such that cyclical variations are
damped.
Oscillatory dynamics is observed when agent’s preferences are highly sensitive to hu-
man capital (high β) and to pollution (low η), hence for high elasticity of environmental
170
4.4 Environmental policy implications
awareness to the green development index. 53 Thus, environmental awareness (γ1) expe-
riences important variations along the converging trajectory, and so do levels of environ-
mental quality and human capital. Some generation experience higher levels and growth
rates of human capital and environmental quality than others. As Seegmuller and Verchère
(2004) show, such cyclical convergence make the welfare varies across generations and
corresponds to intergenerational inequalities. 54 This result emphasizes that taking into ac-
count the short run is important to address environmental policy issue.
Our results in transitional dynamics are opposed to Zhang (1999), who finds that gree-
ner preferences are necessary to avoid complicated dynamic structure. 55 Instead, they are
close to Ono (2003b), who argues that concerns for the environment would cause oscilla-
tions. However, his mechanism goes through innovation and corresponds to higher levels
of exogenous green preferences, while in our setup such dynamics arises from the endoge-
nization of environmental awareness and the feedback effect of environment and education
on green behaviors.
4.4 Environmental policy implications
In this section, we analyze the implications of environmental policy. More precisely,
we attempt to emphasize the impact of policy on intergenerational inequalities in the short-
run and on the long-term growth rate.
53. Indeed, the elasticity is defined as (η−β)X(1+X)(β+ηX) . At a given X , oscillations occurs for high elasticity.
54. Such complex dynamics can also illustrate economic volatility, as Varvarigos (2011) argues.55. He develops a model à la John and Pecchenino (1994) to study the cases of nonlinear dynamics and
endogenous fluctuations.
171
4 Environmental policy and growth in a model with endogenous environmental awareness
4.4.1 The short-term effect of environmental tax
We point out, previously, that the economy may exhibit complex dynamics when en-
vironmental awareness is endogenous. We wonder then how a tighter environmental tax
affects this short-term situation and if the use of a policy mix allows to reduce intergene-
rational inequalities. Focusing on the BGP in the regime with private maintenance, where
damped oscillations may occur, we examine the effect of an increase in environmental tax
on transitional dynamics :
Proposition 21. Under Assumption 7 and β > η, when there is private maintenance at the
stable BGP, an increase in τ implies that :
— If the BGP remains in the regime with private maintenance (σ < σ(τ)), there exists
a σ ∈ (0, 1) such that :
— For σ < σ, oscillations are less frequent.
— For σ > σ, oscillations may be more or less frequent.
— If the BGP moves to the regime without private maintenance (σ > σ(τ)), there is no
damped oscillations.
Proof. See Appendix 4.6.4.
From Proposition 21, we highlight that the government intervention may neutralize
or generate damped oscillations. It comes from the fact that policy shapes the trade-off
between maintenance and education spendings, and hence the mechanism driving oscilla-
tions.
When σ < minσ; σ(τ), an increase in environmental tax allows to reduce the case
where oscillations arise. As highlighted in Section 4.3, cyclical convergence may occur
172
4.4 Environmental policy implications
since green preferences are endogenous. In this situation, along the convergence path, edu-
cation variations stem from several effects through Q, γ1 and h. The former effects drive
oscillations while the latter works in the reverse. As long as σ is low enough, a tighter tax
reinforces the impact of human capital on private education spending. Indeed, the fall in
wage, entailed by tax, is overcompensated by the increase in education subsidy. Education
spending is mainly driven by the government’s action and hence become less sensitive to
γ1. As a result, oscillatory trajectories are less frequent, i.e. the condition to observe dam-
ped oscillations is stricter.
When σ < σ < σ(τ), policy may increase oscillation cases. The intuition is the fol-
lowing. Public revenue devoted to abatement σ is not high enough to be at the regime
without private maintenance (where there is no fluctuation), but at the same time, is too
high to ensure that the amount of education subsidy prevents oscillations. Indeed, environ-
mental tax diminishes the influence of human capital on private education spending and
oscillations may occur more frequently. Specifically, the necessary condition to observe
damped oscillations is more easily satisfied.
In the previous section, we emphasize that an increase in environmental tax makes the
regime without private maintenance (npm) more frequent (σ(τ) goes down) and that the
convergence to the BGP in this regime is always monotonous. Thus, the government may
also avoid intergenerational inequalities by fixing a sufficiently high tax on pollution and
devoting a large share of public spending to maintenance. For a BGP initially in the regime
with private maintenance (pm), if σ(τ) becomes lower than σ, the BGP moves to the re-
gime without private maintenance, there is no more trade-off between private choices and
hence transitional dynamics does not result in oscillations.
Thus, from Proposition 21, we point out that with endogenous concerns, the environ-
173
4 Environmental policy and growth in a model with endogenous environmental awareness
mental tax may favor intergenerational inequalities around the BGP when the product of
the tax is not correctly used. The implications of environmental policy in the short-run
are rarely studied. A notable exception is Ono (2003b), who emphasizes that a sufficient
increase in environmental tax may shift the economy from a fluctuating regime to a BGP
where capital and environmental quality go up perpetually. In this paper, we highlight that
the use of tax revenue is decisive for such a change.
4.4.2 The long-term effect of environmental tax
In accordance with the concept of sustainable development 56, the environment is also
and above all a long-term concern. In this respect, we want to study what solutions the
government can put in place to achieve higher long-term growth of both human capital
and environmental quality. Thus, we examine how policy affects the stable BGP and the
corresponding growth rate.
Using equations (4.4) and (4.12), the expression of the long-term growth rate is given
by :
1+gH =
ǫ
[
γ2µ[A(1−τ)c3+ε2((1−α)Xpm+AN(bστ−a))]+(1−σ)τA(γ1c1+c3(1+γ2µ))
γ1c1+(1+γ2µ)c3
]µ
pm
ǫ[
A[γ2µ(1−τ)+τ(1−σ)(1+γ2µ)]1+γ2µ
]µ
npm
(4.16)
A tighter environmental policy influences the long-term growth rate through several chan-
nels. First directly, by affecting the trade-off between education and maintenance acti-
vities. Second indirectly, by modifying the green development index and environmental
56. The Brundtland Commission (WCED, 1987) defines the sustainable development as “developmentthat meets the needs of the present without compromising the ability of future generations to meet their ownneeds” .
174
4.4 Environmental policy implications
preferences. 57 Therefore, the global impact is ambiguous.
In the following proposition, we emphasize how authorities can improve the growth
rate along the stable BGP :
Proposition 22. Under Assumptions 7 and β > η, following an increase in τ :
— When the BGP remains in the pm regime (σ < σ(τ)), there exists an interval
(σ(τ), σ(τ)) such that the growth rate goes up for σ(τ) < σ < σ(τ).
— When the BGP is initially in or moves to the npm regime (σ > σ(τ)), the growth
rate is enhanced and is higher than in the pm regime for σ < 11+γ2µ
.
Proof. See Appendix 4.6.5
Considering a tighter tax, an economy initially with private maintenance may switch
to the other regime. When the BGP remains in the regime with private maintenance (σ <
σ(τ)), an increase in tax favors both human capital and environmental quality if the allo-
cation between public spendings σ is intermediary. The reason is that extreme allocation
makes one private spending too expensive relatively to the other, which leads agents to
neglect one of the two actions driving growth. When σ is too low, policy favors mainly
education spending. Despite the improvement in γ1 it entails, the increase in tax makes
the private maintenance too expensive, such that the environment deteriorates, and so does
the growth rate. Conversely, when σ is too high, the tax revenue contributes mostly to pu-
blic maintenance. Even if the private investment in environment diminishes (in favor of
57. Note that examining the impact of environmental awareness component on growth, we obtain that thestronger environmental concerns, the lower the growth rate, as in ?. In their paper, a raise in environmen-tal awareness always reduces physical capital accumulation. Here, γ1 affects growth trough an additionalchannel as the environment improves education. Nevertheless, the negative direct impact of environmentalawareness on education more than offsets the improvement of the environment. The proof is available uponrequest.
175
4 Environmental policy and growth in a model with endogenous environmental awareness
education spending), education cost is too high, which weakens human capital accumu-
lation. Therefore, authorities can increase growth by providing support to both education
and maintenance.
When the BGP is or moves to the regime without private maintenance (σ > σ(τ)), 58
a tighter tax leads to the highest growth rate as long as it is accompanied by a sufficient
support for human capital (σ < 1/(1 + γ2µ)). The intuition is the following. On one
hand, in the npm regime, maintenance is entirely public despite the willingness of agent
to contribute privatively to pollution abatement. Therefore, the environmental quality is
sufficiently good at this regime. On the other hand, when the government allocates also a
high enough share of its budget to education, the negative effect of tax on available income
is more than offset by the positive effect through education subsidy. In this way, human
capital accumulation and the environment are enhanced.
Note that when σ > 1/(1+γ2µ), a regime switch can be growth-reducing, particularly
if the increase in τ moving the BGP to the regime without private maintenance is im-
portant. In this case, the negative income effect exceeds the positive impact of education
subsidy and hence human capital accumulation deteriorates.
Our results contribute to the literature on environmental policy and growth. As in
Ono (2003a;b), environmental taxation exerts competing effects on the long-run econo-
mic growth. However, while he observes a positive relationship between the tax and the
long-term growth only for intermediary level of the tax rate, we emphasize that a tighter
policy is always growth-enhancing as long as the tax revenue is well allocated. The re-
lationship that we observe between tax and long-term growth rate is also tied to results
obtained in a recent branch of this literature, that considers the role of human capital. In a
model with a R&D sector reducing pollution, Grimaud and Tournemaine (2007) find that a
58. Following an increase in τ , a BGP initially in the npm regime cannot move to the pm regime
176
4.4 Environmental policy implications
higher environmental tax decreases the price of education relatively to the polluting good,
such that a tighter tax always promotes growth. The same link is present in Pautrel (2011),
Pautel (2012) when lifetime is finite. In the first paper, this holds when abatement sector
is more human capital intensive than final output sector, while in the second this is due to
the fact that pollution stems from physical capital. In these three studies, the underlying
mechanism is that an increase in tax favors education at expense of polluting activities. We
differ from them by pointing out the role of policy mix. The government can implement a
win-win policy only if it allocates properly the revenue of the environmental tax between
education subsidy and public pollution abatement.
We emphasize that the recycling of environmental tax can have double dividend : en-
vironmental quality improvement and economic benefit. However, we differ from the li-
terature on the “double dividend hypothesis” which considers the economic dividend as
an improvement in economic efficiency from the use of environmental tax revenue to re-
duce other distortionary taxes (e.g Goulder, 1995, Chiroleu-Assouline and Fodha, 2006).
Instead, we emphasize that, when the tax on polluting activities is recycled in public main-
tenance and education subsidy, policy can promote both environmental quality and human
capital.
4.4.3 How the government policy can improve the short-and long-term situations ?
We emphasize previously that, in the short run the government should intervene to
avoid intergenerational inequalities, while in the long run it seeks to enhance growth. In
this section, we examine how a tighter tax can satisfy these two objectives. By considering
the properties of the function σ(τ), that defines the regimes with and without private main-
tenance, with the properties of σMin(τ) and σMax(τ), that determine the possible policy
177
4 Environmental policy and growth in a model with endogenous environmental awareness
schemes (see Assumption 7), we summarize the results of Propositions 21 and 22 :
178
4.4 Environmental policy implications
FIGURE 4.5 – Short-and long-term implications of a tighter taxat σ given, when a
b< 1
1+γ2µ
This figure depicts the implications of a tighter tax for a given σ. 59 Fourth areas are dis-
tinguished.
In areas 1 and 2, the BGP equilibrium is in the regime with private maintenance. Fo-
cusing on the short run, a tighter tax makes the occurrence of oscillations less likely when
σ < σ (area 1), while the opposite holds when σ > σ (area 2). In these areas, to achieve
the long-term objective, an increase in the pollution tax has to be associated with an in-
termediary level of σ. Unfortunately, no clear conclusion emerges when comparing these
values of σ with the one corresponding to benefit in the short-run.
As long as the tax rate is sufficiently high, the economy can achieve areas 3 and 4,
where the BGP is in the regime without private maintenance and hence contributions to the
pollution abatement are entirely public. At this state, the short-term issue vanishes and the
59. From Proposition 22, we deduce that a growth enhancing policy exists in the regime npm if and onlyif a
b< 1
1+γ2µ, where the ratio a
brepresents the minimum level of the threshold σ(τ) over which the economy
is in the regime without private maintenance.
179
4 Environmental policy and growth in a model with endogenous environmental awareness
economy performs a higher long-term growth rate when education subsidy is sufficiently
high (σ low, area 4). Furthermore, as stressed in Proposition 22, for a given σ, the growth
rate is higher in area 4 than in the other regime.
As a result, we identify the most favorable tax scheme as the one where the environ-
mental protection results exclusively in public spending and education support is suffi-
ciently high (area 4). In such a way, there is no more trade-off between education and
maintenance, which would lead eventually to damped oscillations and hence to inequa-
lities. Moreover, since public maintenance and education subsidy are sufficiently high,
human capital accumulation is favored without damaging the environment.
Note that the design of this policy depends on environmental awareness. For a given
σ, the higher green preference parameters, the higher the tax rate required to achieve the
area 4 (the curve σ(τ) moves to the right).
4.5 Conclusion
In this paper, we examine the implications of environmental policy on growth when en-
vironmental awareness is endogenously determined by both education and pollution. The
government can strengthen its environmental commitment by increasing a tax on pollu-
tion and allocate tax revenue between two categories of environmental expenditure : direct
one with public pollution abatement and more indirect one through education subsidy. We
show that there exists a unique positive BGP, characterized by sustainable development. It
can be either in a regime with private environmental maintenance or in a regime with only
public environmental maintenance, depending on the design of environmental policy.
When the BGP is with private maintenance, we reveal that the economy may display
damped oscillations for highly sensitive environmental preferences. These oscillations are
180
4.5 Conclusion
due to the feedback effect of human capital and environmental quality on endogenous
green preferences, which shapes the trade-off between private choices. Such complex dy-
namics makes the welfare vary across generations, and entails intergenerational inequa-
lities. Conversely, when the BGP is in the regime with only public maintenance, there is
no more private choices in maintenance and hence damped oscillations do not occur. In
addition, we prove that the highest growth rate is achieved in this regime when the share
of public spending devoted to education is sufficiently high. In this case, human capital
accumulation is favored without damaging the environment.
As a result, we reveal that environmental policy plays a crucial role in avoiding interge-
nerational inequalities and in improving growth of both human capital and environmental
quality. More precisely, we conclude in favor of policy mix and underline that the most
favorable policy scheme corresponds to an intermediary allocation of budget between pu-
blic environmental maintenance and education.
Given the varieties of externalities considered in this model, it would be interesting to
provide some insights about the optimal policy scheme that a government should imple-
mented to maximize the intertemporal welfare. Our setting does not allow to conduct a
welfare analysis, but positive externalities in education and in environment suggest that a
mix of instruments would be required to decentralize the optimal policy.
181
4 Environmental policy and growth in a model with endogenous environmental awareness
4.6 Appendix
4.6.1 Equilibrium
First Order Conditions
The maximization of the consumer program (4.6) leads to the following first order
conditions on education expenditure and on environmental maintenance :
∂U
∂et= 0 ⇔ 1− θet
ct=γ2µ
et(4.17)
∂U
∂mt
6 0 ⇔ 1
ct>
γ1t(ε1 + ε2b)
ε1mt + ε2Qt+1
(4.18)
From equations (4.3), (4.5) and the first order conditions (4.17) and (4.18), we deduce
the optimal choices in terms of education and maintenance in two regimes : an interior
solution, where individuals invest in environmental protection mt > 0 (hereafter pm) and
a corner solution without private contribution to the environment mt = 0 (hereafter npm).
et =
(
γ2µ1−θet
)(
(ε1+ε2b)wtht+ε2[(1−α)Qt−aYt+b(Mt+NGmt )]
(1+γ1t+γ2µ)(ε1+ε2b)
)
pm
wthtγ2µ(1+γ2µ)(1−θet )
npm
mt =
γ1t(ε1+ε2b)wtht−ε2(1+γ2µ)[(1−α)Qt−aANht+b(Mt+NGmt )]
(1+γ1t+γ2µ)(ε1+ε2b)pm
0 npm
182
4.6 Appendix
At the symmetric equilibrium, Mt = mt(N − 1), the wage equilibrium is wt = A(1− τ),
the production function is Yt = ANht and the government budget constraint is given by
(4.9). The Nash intertemporal equilibria are thus given by :
et =
γ2µ[c3A(1−τ)ht+ε2((1−α)Qt+ANht(bστ−a))]γ1tc1+(1+γ2µ)c3
+ (1− σ)τAht pm
Aht[γ2µ(1−τ)+τ(1−σ)(1+γ2µ)]1+γ2µ
npm
mt =
γ1tc1Aht(1−τ)−ε2(1+γ2µ)[(1−α)Qt+ANht(bστ−a)]γ1tc1+(1+γ2µ)c3
pm
0 npm
Condition for the regime without private maintenance
The condition (4.14) can be written as :
P(Xt) ≡ X2t ε2(1 + γ2µ)(1− α) +Xt[ε2(1 + γ2µ)(1− α+AN(bστ − a))−Aηc1(1− τ)]
−A[βc1(1− τ)− (bστ − a)Nε2(1 + γ2µ)] > 0
(4.19)
For Xt ∈ [0,+∞[, the polynomial P(Xt) = 0 admits one solution or none solution.
— When, βc1(1− τ)− (bστ − a)Nε2(1+ γ2µ) > 0 the polynomial P(Xt) = 0 admits
one solution Xt = Λ > 0.
— When βc1(1− τ)− (bστ − a)Nε2(1 + γ2µ) 6 0, the polynomial is positive or nul
∀ Xt > 0. In this case, the economy is always in the corner regime (i.e Λ = 0).
As a result, the economy is in the corner regime when Xt > Λ, with Λ > 0.
183
4 Environmental policy and growth in a model with endogenous environmental awareness
Growth rates
We define the endogenous growth rate of human capital gH , and environmental quality
gQ, using equations (4.4), (4.5), (4.12) and (4.11) :
1+gHt =
ǫ[
γ2µ[A(1−τ)c3+ε2((1−α)Xt+AN(bστ−a))]+(1−σ)τA(γ1tc1+c3(1+γ2µ))γ1tc1+(1+γ2µ)c3
]µ
pm
ǫ[
A[γ2µ(1−τ)+τ(1−σ)(1+γ2µ)]1+γ2µ
]µ
npm
(4.20)
In the npm regime, human capital growth rate is constant, as education spending does not
depend on the environment. The pm regime is characterized by a human capital growth rate
increasing in the green development index, directly and indirectly though environmental
awareness γ1t.
1 + gQt =
(1−α)Xt(γ1tc1+c2)+AN [γ1t c1(1−τ)b+(γ1tc1+c2)(bτσ−a)]Xt(γ1tc1+c3(1+γ2µ))
pm
1− α + AN(bστ−a)Xt
npm
(4.21)
In the case with private maintenance, the green development indexXt has a direct negative
impact on the growth rate of environmental quality and an indirect positive effect through
environmental awareness γ1t. In the corner solution, gQt is always negative without public
abatement. 60 However, when the government intervenes, the growth of the environmental
quality at the corner may be positive for sufficiently high share of policy devoted to public
environmental maintenance (σ).
60. When mt = 0 and σ = 0, gQt is increasing in Xt : an increase in Xt fits in with a human capitaldecline and hence with a lower pollution.
184
4.6 Appendix
4.6.2 Proof of Proposition 19
Properties of the dynamical equation
From Definitions 5 and 6 and equations (4.13) and (4.14), we emphasize the properties
of the dynamic equation characterizing equilibrium paths, F(Xt), defined on (0; +∞) :
— When Xt ∈ (0 ; Λ) the function is given by equation (4.15 npm). We have
F(0) =AN(β c1(1− τ)b+ (βc1 + c2)(bτσ − a))
ǫ [Aγ2µ ( c3(1− τ) + ε2N(bστ − a)) + (1− σ)τA(β c1 + (1 + γ2µ)c3)]µ [β c1 + (1 + γ2µ)c3]
1−µ
Finally, with equation (4.14), limXt→Λ−
F(Xt) =(1−α)Λ+AN(bστ−a)
ǫ[
γ2µA(1−τ)+(1−σ)τA(1+γ2µ)1+γ2µ
]µ ≡ v.
— When Xt ∈ [Λ ; +∞), the function is given by equation (4.15 pm). F is increasing
and linear in X , F(Λ) = v and limX→+∞ F(X) = +∞.
As limXt→Λ
F(Xt) = F(Λ), the function is continue on (0; +∞).
Existence and Unicity of the Balanced Growth Path
npm solution. A BGP in the npm regime is characterized by Xnpm = F(Xnpm).
Using (4.15 npm), we obtain :
Xnpm =AN(bτσ − a)
ǫ[
A[γ2µ(1−τ)+(1−σ)τ(1+γ2µ)]1+γ2µ
]µ
− (1− α)(4.22)
To exists Xnpm has to be positive. Following (4.20 npm), if the denominator of (4.22) is
negative, so does the growth rate in the npm regime. Therefore, the case where the deno-
minator and the numerator of (4.22) are negative is meaningless. The existence conditions
are thus σ > abτ
and A1 > 0 with A1 ≡[
A[γ2µ(1−τ)+(1−σ)τ(1+γ2µ)]1+γ2µ
]µ
− (1 − α). Note that
the condition A1 > 0 implies that limX→+∞
F(X)X
< 1.
185
4 Environmental policy and growth in a model with endogenous environmental awareness
Then, we have to check the admissibility of the steady state, i.e. if it effectively belongs
to the npm region. To do this, we examine the sign of F(Λ)−Λ. Under condition A1 > 0
and σ > abτ
, Xnpm is admissible if F(Λ) ≥ Λ, which is equivalent to Xnpm ≥ Λ.
pm solution. A BGP in the pm regime is characterized by Xpm = F(Xpm). As we fo-
cus on X > 0, we determine the solutions Xpm which satisfy F(Xpm)/Xpm = 1. Using
equations (4.13) and (4.15 npm.), it corresponds to (for the sake of simplicity, subscripts
on X are removed) :
ǫ[γ2µ(Ac3(1−τ)+ε2((1−α)X+AN(bτσ−a)))+Aτ(1−σ)(c1 β+ηX
1+X+(1+γ2µ)c3)]
µ
[β+ηX1+X
c1+c3(1+γ2µ)]1−µ
= (1− α)(
β+ηX1+X
c1 + c2)
+AN(β+ηX
1+Xc1(b(1−τ)+bτσ−a)−c2(a−bτσ))
X
We define D1(X) and D2(X), respectively the term on the left and on the right hand side.
Their properties are :
— D1 is decreasing and then increasing with X . Moreover, D1(X) > 0 for all X ,
limX→0
D1(X) = C > 0, with C a constant and limX→+∞
D1(X) = +∞.
— Concerning D2 we have that if A2 ≡ βc1(b(1−τ)+bτσ−a)−c2(a−bτσ) > 0 (resp.
< 0), limX→0
D2(X) > 0 (resp.< 0). Moreover, limX→+∞
D2(X) = (1−α)(ηc1+c2) > 0.
The condition A2 > 0 guarantees that the curves D1(X) and D2(X) cross at least once in
the positive area. Thus, a positive solution exists. From equation (4.15 pm.) and A2 > 0,
we have dF(Xpm)/dX < 1, hence the positive solution Xpm is unique.
We check the admissibility of the steady state, i.e. if it effectively belongs to the pm
region. Under condition A2 > 0, Xpm is admissible if limX→Λ
F(X) < Λ, which is equivalent
to Xpm < Λ. As Xpm 6 Xnpm, the condition Xnpm < Λ guarantees that Xpm is admissible.
186
4.6 Appendix
The case where both Xpm and Xnpm are admissible is excluded as long as A1,A2 >
0 and σ > abτ
. When Xpm exists (condition A2 > 0 verified) and is admissible, then
F(Λ) < Λ. But under condition A1 > 0 the slope of F in the npm regime is lower than
one, which means that F cannot cut the bissectrice in this regime. Reversely, when Xnpm
exists (conditions A1 > 0 and σ > abτ
verified) and is admissible then F(Λ) > Λ. Under
condition A2 > 0, this implies that F does not cut the bissectrice in the pm area.
To sum up, under these conditions it is not possible to have Xpm < Λ 6 Xnpm, and
hence to have cases with multiple steady states : when Xnpm < Λ, the BGP is in the regime
pm, while when Xnpm > Λ, the BGP is in the regime npm.
For the rest of the paper, we rewrite the conditions A1,A2 > 0, as σMin(τ) < σ <
σMax(τ) with
σMin(τ) ≡a(βc1 + c2)− bβc1(1− τ)
bτ(βc1 + c2)and σMax(τ) ≡
A(γ2µ+ τ)− ǫ−1/µ(1 + γ2µ)(1− α)1/µ
Aτ(1 + γ2µ)
σMin(τ) is increasing in τ from limτ→0 σMin = −∞ to limτ→1 σMin = a/b, while
σMax(τ) is decreasing in τ from limτ→0 σMax = +∞ to limτ→0 σMax = 1− 1ǫ1/µA
.
Condition for the nature of the BGP. The analysis of admissibility conditions come down
to Xnpm < or > Λ.
From equation (4.19) in Appendix 4.6.1, Xnpm > Λ is equivalent to P(Xnpm) > 0.
We define P(Xnpm) ≡ J (τ, σ), where the function J is increasing in τ and σ. Under
A1 > 0 :
— For τ = 0, Xnpm < 0, J < 0 and does no longer depend on σ.
— For τ = 1, we get J > 0 ∀ σ ∈ [0, 1].
We depict a representation of J at given τ :
187
4 Environmental policy and growth in a model with endogenous environmental awarenessJ (σ, τ)
0
τ = 1
τ ∈ (0, 1)
τ = 0
σσc(τ)
FIGURE 4.6 – Function J at given τ
We deduce that there exists a σc(τ) decreasing in τ such that J = 0, with limτ→0
σc(τ) =
+∞ and σc(1) < 0. Thus a minimum level of tax is required to make the npm regime
possible. When σ < σc(τ), we get P(Xnpm) < 0, meaning that the equilibrium is in
the pm regime. Respectively, when σ > σc(τ) we get P(Xnpm) > 0, and from equa-
tions (4.19) and (4.22) we have Xnpm > Λ if and only if σ > abτ
. Thus, the BGP is
in the npm regime when σ > Maxσc(τ); abτ ≡ σ(τ) and in the pm regime when
σ < σ(τ). When σ = σ(τ), the BGP is in the npm regime if σ(τ) = σc(τ) and in the
pm regime if σ(τ) = abτ
. Note that σ(τ) > σMin(τ) and limτ→1 σ(τ) < limτ→1 σMax(τ).
4.6.3 Proof of Proposition 20
pm solution. We use the scaling parameter ǫ in order to normalize the steady state Xpm
to one. There is a unique solution ǫ∗ such that Xpm = 1 and from equation (4.15 npm.),
the expression of the normalization constant is given by :
ǫ∗(Xpm) ≡[
(1− α) Xpm [γ1c1 + c2] +AN(γ1c1b(1− τ) + (γ1c1 + c2)(bτσ − a))]
×[
γ2µA c3(1− τ) + γ2µε2[
(1− α)Xpm +AN(bστ − a)]
+ (1− σ)τA(γ1 c1 + (1 + γ2µ)c3)]−µ ×
[γ1 c1 + (1 + γ2µ)c3]µ−1
188
4.6 Appendix
Then, by differentiating equation (4.15 npm.) and analyzing it around the steady state
Xpm = 1 and ǫ ≡ ǫ∗(Xpm), we obtain :
dXt+1 =((1−α)(γ1c1+c2+γ1′c1)+ANγ1′c1(b(1−τ)+bτσ−a))B1B2−B3(µB2(γ2µε2(1−α)+(1−σ)Aτγ1′c1)+(1−µ)γ′
1c1B1)B3B1B2
dXt
(4.23)
with γ1 =β+η2, γ1
′ = (η−β)/4, B1 = γ2µA c3(1−τ)+γ2µε2 [(1− α) + AN(bστ − a)]+
(1 − σ)τA(γ1 c1 + (1 + γ2µ)c3), B2 = γ1 c1 + c3(1 + γ2µ) and B3 = (1 − α)(γ1 c1 +
c2) + AN(γ1c1b(1− τ) + (γ1c1 + c2)(bτσ − a)).
From (4.23), we get dXt+1/dXt < 1. Thus, when dXt+1/dXt > 0, transitional dyna-
mics is monotonous and the BGP equilibrium is locally stable. Using equation (4.23), we
have dXt+1/dXt > 0 if and only if :
B2(1− α) [(1− σ)τAB2 + γ2µ(Aε1(1− τ) + ε2(1− µ))] (γ1c1 + c2)
+ B2(1− α)Aε2bN(1− τ)(γ1c1(1− µ) + c2)
+ γ1′c1B5 [γ2µ(1− α+AN(b− a) + bNτ(σ − 1))(bN(1 + γ2µ)ε2(1− µ) + µB2)]
+ γ1′c1B5ANb(1 + γ2µ)(γ2µε1(1− µ) + (1− σ)τB2) > 0
with B4 ≡ 1−α+AN(b(1− τ) + bτσ− a) and B5 ≡ Ac3(1− τ) + ε2(1−α+AN(bτσ− a)).
Rewriting this expression, we have dXt+1/dXt > 0 if and only if the following polynomial
is positive :
R(β) ≡ a1β3 + a2β
2 + a3β + a4
with a4 > 0 and expressions for a1, a2 and a3 given by :
a1 =c318(1− σ)(1− α)τA > 0
189
4 Environmental policy and growth in a model with endogenous environmental awareness
a2 =c21(1−α)
4
[
γ2µ(1− µ)B5 + (1− σ)τA[2c3(1 + γ2µ) + c2] + γ2µ2ε1A(1− τ)
]
+c218
[
3ηc1(1− α)(1− σ)τA− B5
(
γ2µ2B4 + bAN(1 + γ2µ)τ(1− σ)
)]
a3 =c31η
2(1−σ)(1−α)τA8 +
2ηc214 (1− α)
[
γ2µ(1− µ)B5 + (1− σ)τA[2c3(1 + γ2µ) + c2] + γ2µ2ε1A(1− τ)
]
+ c1(1−α)2
[
3(1− σ)τAc2c3(1 + γ2µ) + (1− µ)γ2µB5(c3(1 + γ2µ) + c2) + 2γ2µ2A(1− τ)ε1c3(1 + γ2µ)
]
− c1B5
4
[
γ2µ2c3(1 + γ2µ)B4 + bN(1 + γ2µ)(γ2µ(1− µ)B5 + (1− σ)τAc3(1 + γ2µ))
]
We have R(1) which is a polynomial of degree three in N . When N = 0, we have
R(1) > 0, while whenN tends to ∞R(1) < 0. As a result, there exists a critical threshold
N over which the dynamics is oscillatory for β = 1. We can thus conclude that forN > N
there exists a β ∈ (0, 1] over which the dynamics is oscillatory.
We examine the stability of the equilibrium when dynamic is oscillatory. From equa-
tion (4.23), we have dXt+1/dXt > −1 if and only if :
[(1− α)(γ1c1 + c2 + γ1′c1) + ANγ1
′c1(b(1− τ) + bτσ − a)]B1B2 + B1B2B3
− [(1− µ)c1γ1′B1 + B2µ (γ2µε2(1− α) + (1− σ)Aτγ1
′c1)]B3 > 0
Replacing expressions B1, B1 and B3, we finally obtain :
c3(1 + γ2µ)c2(1− α)(B6 + ε2bNµ) + γ2µ(γ1c1)2(B6(1− α) + B5B4)
+ γ1c1(1 + γ2µ)[(1− α)ε1(B6 + ε2bNµ) + γ2µε1(1− α+AN(bτσ − a)) + c3(B4 + B6)]
+ γ2µγ1′c1(1 + γ2µ)(B4c3 − (1− µ)ε1(1− α+AN(bτσ − a)))
+ γ2µ2γ1
′γ1c21B5B4 + B5c3(1 + γ2µ)c2(1− α+AN(bτσ − a))
+ (1− σ)τA(γ1c1 + c3(1 + γ2µ))γ′1c1bN(1 + γ2µ)B5
+ (1− σ)τA(γ1c1 + c3(1 + γ2µ))2[γ1c1B4 + c2(1− α+AN(bτσ − a)) + (1− α)(γ1c1 + c2)]
with B6 = c2 + ε2bN(1 + γ2µ) > 0.
As −γ1′ < γ1 and c3B4 > bNB5, we easily see that this term is always positive. The
190
4.6 Appendix
positive BGP equilibrium is always locally stable.
npm solution. The npm BGP is obtain from (4.15 pm) and given in Appendix 4.6.2.
We differentiate equation (4.15 pm) and obtain :
dF(Xt)
dXt
=(1− α)
ǫ[
A[γ2µ(1−τ)+(1−σ)τ(1+γ2µ)]1+γ2µ
]µ
Under Assumption 7, the slope of F(Xt) in the npm regime is always positive and lower
than one, the npm BGP is thus monotonously stable.
4.6.4 Proof of Proposition 21
The condition to observe oscillatory cases (i.eN > N ) depends on policy instruments.
Moreover, by examining the terms a2 and a3 given Appendix 4.6.3, we conclude that if a2
is positive, a3 is positive as well. A necessary condition is thus required to have N > N .
The term a2 given in Appendix 4.6.3 has to be negative. An analysis of the impact of the
policy instruments on the polynomial R(β) is not analytically tractable, but the analysis
of sgn
∂a2∂τ
gives us interesting intuitions. Using the expression of a2 given in Appendix
4.6.3 we obtain :
sgn
∂a2∂τ
= −(1− σ)bAN(1 + γ2µ(1− µ))(S2 − τAε2bN(1− σ))(τ(1− σ)S3 + (1− τ)γ2µAε1)
+(1− σ)(bτ(1− σ)AN(1 + γ2µ(1− µ)) + S1)(ε2ANbS4 + S3S2) + 3(1− σ)ηc1AS4
−Aε1µ3(1− α+AN(b− a))(S1 + bτ(1− σ)AN(1 + γ2µ(1− µ)))
with S1 = µ2γ2(1 − α + AN(b − a)), S2 = A(1 − τ)ε1 + ε2(AN(b − a) + 1 − α), S3 =
A(3ε1(1+γ2µ)+2ε2bN+γ2µε2bN(1+µ)) and S4 = γ2µ(ε2(1−µ)(1−α+AN(b−a))+Aε1)
We can define sgn
∂a2∂τ
as a polynomial of degree three in σ, with ∂a2∂τ
< 0 when
191
4 Environmental policy and growth in a model with endogenous environmental awareness
σ = 1 and ∂a2∂τ
> 0 when σ = 0. Since sgn
∂a2∂τ
is decreasing in σ for σ ∈ [0, 1], there
exists a critical value σ ∈ (0, 1) such that : for 0 < σ < σ, ∂a2∂τ
> 0 and for σ < σ < 1,
∂a2∂τ
< 0. Moreover, from Assumption 7, limτ→0 σMin(τ) < σ < limτ→1 σMax(τ). When
σ is sufficiently low, a tighter tax tightens the condition to observe oscillatory cases, while
when σ is high enough the condition to observe oscillatory dynamics may relaxed.
4.6.5 Proof of Proposition 22
We examine the impact of taxation on the growth rate along the BGP.
pm solution. Using equation (4.20 npm) with Xt = Xpm we have :
sgn(
∂gpm∂τ
)
= V2
(
(1− σ)(γ1c1 + c3)− σγ2µε1 + γ2µε2(1− α)∂Xpm
∂τ
)
+ c1(β−η)
(1+Xpm)2∂Xpm
∂τγ2µ (V1 − τ(1− σ)AV2)
with V1 = γ2µAc3(1−τ)+γ2µε2[
(1− α)Xpm +AN(bστ − a)]
+(1−σ)τA(γ1c1+(1+γ2µ)c3)
and
V2 = γ1c1+(1+γ2µ)c3. From the implicit function theorem and equation (4.15), we have :
∂Xpm
∂τ= V2XpmA[(γ1c1(σ−1)+σc2)V1N−µV3(−σε1γ2µ+(1−σ)(γ1c1+c3))]
c1Xpm(β−η)
(1+Xpm)2(V1V2(Xpm(1−α)+AN(b(1−τ)+bτσ−a))−µV2V3(1−σ)τA+(1−µ)V1V3)+µV2V3Xpmγ2µε2(1−α)+ANV1V2V4
with V3 = (1 − α)Xpm [γ1c1 + c2] + AN [γ1c1b(1− τ) + (γ1c1 + c2)(bτσ − a)] and V4 =
c2(bτσ − a) + γ1c1b(1− τ(1− σ))
192
4.6 Appendix
Thus, substituting ∂Xpm
∂τin sgn
(
∂gpm∂τ
)
, we finally obtain :
sgn(
∂gpm∂τ
)
=(
γ2µε2(1− α)V2XpmAN +c1XpmAN(β−η)
(1+Xpm)2[V1 − τ(1− σ)AV2]
)
(γ1c1(σ − 1) + σc2)
+(
c1Xpm(β−η)
(1+Xpm)2
[
V2(Xpm(1− α) +AN(b(1− τ) + bτσ − a)) + V3
]
+ V2V4AN)
(−σε1γ2µ+ (1− σ)(γ1c1 + c3))V1
Under Assumption 7, policy improves the BGP growth rate when the following sufficient
condition is satisfied :
f1(σ) < σ < f2(σ)
with f1(σ) ≡ γ1c1γ1c1+c2
< 1 and f2(σ) ≡ γ1c1+c3γ1c1+c2+γ2µε1
< 1. These two functions are increa-
sing in γ1, and as ∂γ1∂Xpm
< 0 and ∂Xpm
∂σ> 0, they are decreasing in σ. As a result, there
exists a unique range of value [σ(τ); σ(τ)] which satisfies this condition. Moreover, under
Assumption 7, limτ→0 σMin(τ) < σ(τ) < σ(τ) < limτ→1 σMax(τ).
npm solution. We use equation (4.20 pm) with Xt = Xnpm and deduce :
sgn(
∂gc∂τ
)
= 1− σ(1 + γ2µ)
A tighter tax is growth promoting as long as σ(τ) < σ < 1/(1 + γ2µ).
Regime switch. We consider the case where an increase in τ leads the economy from a
pm regime to a npm regime. The opposite switch cannot be observed as σ(τ) is decreasing
in τ . For a given σ, we compare equations (4.20 npm) and (4.20 pm), by considering a
higher tax rate in the npm regime (τN ) than in the pm one (τP ). The growth rate in the pm
193
4 Environmental policy and growth in a model with endogenous environmental awareness
regime is higher than in the npm if and only if :
γ2µ(1 + γ2µ)(
c3A(τN − τP ) + ε2((1− α)Xpm + AN(σbτP − a)))
−(1− σ)A(γ1c1 + (1 + γ2µ)c3)(τN − τP )− Aγ2µ(1− τN)γ1c1 > 0
This expression is increasing in σ and from (4.14) is never satisfied when σ = 1/(1+γ2µ).
And according to Appendix 4.6.2, the npm regime exists only if ab< σ. Thus, for a given
σ ∈ (a/b ; 1/(1 + γ2µ)), the growth rate in the npm regime is higher than in the pm one.
Moreover, under Assumption 7, σMin(τ) < 1/(1 + γ2µ) < σMax(τ).
194
General conclusion
This thesis presents a purely theoretical analysis on education and growth. The litera-
ture examining the role of human capital in growth is well documented by empirical and
theoretical studies and provides relevant information to guide the public intervention in
education. Nonetheless, some major properties are ignored, such as the different uses of
human capital in the sectors of the economy, while being important to better identify the
channels through which education spending affects growth. Moreover, while the link bet-
ween growth and environment represents a major challenge, the studies combing growth,
education policy and pollution issues are sparse. The present thesis contributes to the large
literature on growth and human capital by proposing new formalizations and different ap-
proaches to investigating the role and the impact of public intervention in education.
The intensity in human and physical capital differs according to the different sectors of
the economy. Nevertheless, the literature has not provided yet an answer on how a factor
intensity reversal could affect the optimal policy scheme. Chapter one attempts to over-
come this lack. We explore the impact of sectoral properties to design optimal education
policy. Using a two-sector framework, we reveal that the valuation of the social planner
for future generations and the agents’ preferences for education and time, determine the
policy implications of a factor intensity reversal. The intuition is that agents and planner’s
preferences determine the target sector that the government should favor in order to correct
market inefficiencies. Hence, the structural changes occurring in this sector will determine
how policy should be adapted.
195
General conclusion
In the first chapter, we investigate whether the government should favor more inten-
sively human or physical capital accumulation. It is shown that, the relative importance
of human vis à vis physical capital accumulation is modified following a factor intensity
reversal. When the investment sector becomes more intensive in human capital while the
consumption sector becomes more intensive in physical capital, optimal policy turns in
favor of education, as long as the social planner gives a sufficient weight to future gene-
rations. At the aggregated level, these sectoral changes are ignored and the policy recom-
mendations can be different. In conclusion, the optimal policy scheme can be misleading
if we do not consider the two-sector dimension.
A possible extension of the analysis proposed in the first chapter is exploring a setting
where the interactions between the two kinds of capital are diverse. Education may gene-
rate positive spillover effects on the formation of physical capital, while physical capital
accumulation may be the source of positive externalities for human capital. Taking into ac-
count these external effects could provide more insights about how the government should
direct its action.
Recent contributions in the literature on growth and human capital highlight that the
way education policy is financed is crucial to examine the relationship between public
education expenditure and growth. Since these studies are developed in one-sector fra-
meworks, the relative price of education spending is implicitly fixed and equal to one.
Chapter two contrasts with these findings, by considering two sectors in the economy. In a
two-sector framework, a public policy shapes the relative price of education services and
hence the relationship between public education expenditure and growth. While agents’
preferences are not important in the previous studies, we reveal that they are a major de-
terminant of the effect of public education on growth. We identify two situations where
196
General conclusion
the agents’ preferences determine the sensitivity of the relative price to the fiscal policy.
First, when a sectoral tax is implemented to finance public policy rather than a tax on the
aggregate production. Second, when factor intensity differs between sector, whatever the
fiscal policy. This result leads to the following important policy implications : the same
public education policy in developed and developing countries will generate different ef-
fects because of cross countries heterogeneity in preferences.
In the third chapter, we further develop the two-sector models to examine the dynamics
of human capital accumulation in an international context. The main motivation comes
from the fact that the literature addressing the impact of economic integration on human
capital accumulation, assumes that the good used to invest in education or in physical ca-
pital is produced by the same sector and is a tradable good. We propose to differentiate
education services and savings by their tradability, assuming that a non-tradable good is
used to invest in human capital. This assumption entails important modifications : on the
methodological side, our approach stresses that non-tradability of a part of the production
plays an important role because it introduces a crucial variable : the relative price of non-
tradable good in terms of a tradable one. This relative price determines the growth rate and
drives the transitional dynamics of the economies.
Our analysis highlights that the long-term impact of economic integration does not
presume of the short-term one when taking into account the relative price adjustments ge-
nerated by economic integration. The model also suggests that tradable TFP differences
between countries are the major determinant of the impact of integration on growth. Fur-
ther research, comprise identifying the factors that may affect TFP in the tradable sector.
Particularly, that would allow examining how governments can intervene to fill the gap
between developed and developing countries.
197
General conclusion
The theoretical results of chapter three might provide some guidance on education po-
licy. In an international context, government measures implemented to boost student mo-
bility could intensify inter-regional spillovers in human capital. Such externalities would
be welcome because they guarantee growth convergence between countries and may re-
duce the adjustment cost of the growth rate at the time of integration. Another avenue for
further work would be to focus on the way efficiency could be restored in this integrated
economy. Within an economic union with inter-regional externalities in education, an op-
timal policy supposes concerted efforts for the different members.
The evolution of environmental challenges during recent decades underlines the ne-
cessity to include an environmental constraint in the economic growth analysis. Chapter
four considers this feature, by developing an endogenous growth model where environ-
mental quality has an impact on welfare. We pay particular attention to agent’s green
behaviors, because the implementation of measures to raise consumer awareness of the
environmental problem is viewed as an important lever to meet the challenge of sustai-
nability. Environmental awareness is assumed to depend on two factors, identified by the
literature as relevant determinants of ecological action : the level of human capital and
the stock of pollution. In this framework, we study the effects of an environmental policy
mix consisting on measures in education, maintenance activities and pollution taxation.
It turns out that, under a reasonable assumption, an appropriate mix of policy leads to a
win-win situation. First, it allows achieving a sustainable balanced growth path where both
the environmental quality index and human capital grow at a higher rate. Secondly, this
policy avoids cyclical convergence, source of intergenerational inequalities, arising from
endogenous preferences.
A natural extension of this thesis would concern the last chapter where the analysis
198
General conclusion
was conducted in a stylized model with only one sector. The models developed in chap-
ter one through three are particularly relevant to treat environmental issues. Firstly, the
two-sector dimension is adapted as some activities use polluting inputs more intensively
than others. The fourth chapter provides some insights about the type of policy that would
allow reducing pollution without compromising economic growth. With a two-sector mo-
del, we could differentiate between green and brown goods and address this question more
precisely. In particular, a two-sector dimension gives the possibility to consider that the
green good is produced with clean inputs but is more expensive than the brown good. Se-
condly, the international context studied in chapter three also provides interesting features
for the environmental perspective and would allow considering the trans-boundary nature
of pollution. The environment is a global issue that requires a common environmental and
energy policy between countries. Thus, considering a two-country framework would allow
exploring international environmental policy and would contribute to the large literature
that attempts to examine how international trade affects the pollution ? The theory predicts
that, following trade openness, a country that is intensive in human capital and that is cha-
racterized by a higher environmental regulation tends to specialize in the production of
green goods. On the contrary, a country that is intensive in physical capital and with low
regulations will produce mainly brown goods. Thus, polluting activities tend to localize in
developing economies.
On the other hand, it is also admitted that trade can favor the diffusion of green tech-
nologies adopted in developed countries. This second aspect is poorly studied in the theo-
retical literature. In this context, it would be relevant to assume that human capital accu-
mulation promotes the adoption and diffusion of greener technologies. This formalization
would allow considering an additional channel through which trade and human capital
199
General conclusion
may affect pollution. Moreover, this setting would be appropriated to explore the impact
of local and global environmental policies. Given the trans-boundary nature of pollution,
we would like to determine what policy should be implemented to compensate the negative
effect of international trade on overall pollution.
200
References
Alonso-Carrera, J., Caballé, J. and Raurich, X. (2011), “Sectoral composition and macroe-
conomic dynamics”, Barcelona GSE Working Paper Series 545.
Andreoni, J. (1990), “Impure altruism and donations to public goods : a theory of warm-
glow giving ?”, Economic Journal, 100(401), 464-477.
Armellini, M. and Basu, P. (2010), “Altruism, education subsidy and growth”, MPRA
working paper number 23653.
Barro, R. and Sala-i-Martin, X. (1999), “Economic Growth”, MIT Press, Cambridge, MA.
Basu, P., and Bhattari, K. (2012), “Government bias in education, schooling attainment,
and long-run growth”, Southern Economic Journal, 79(1), 127-143.
Becker, G.S. (1964), “Human capital, A theoretical and empirical analysis with special
reference to education”, The University of Chicago Press.
Becker, G.S. (1991), “A treatise on the family”, enlarged edition. Cambridge Mass. Har-
vard University Press.
Benabib, J., and Spiegel, M. (1994),“The role of human capital in economic development :
Evidence from aggregate cross-country data”, Journal of Monetary Economics, 34(2),
143–173.
Berka, M. and Devereux, M.B. (2013), “What determines european real exchange rates”,
Economic Policy 28, 193-242 .
Blackburn, K. and Cipriani, G.P. (2002) “A model of longevity, fertility and growth”, Jour-
nal of Economic Dynamics and Control, 26(2), 187-204.
201
References
Blankenau, F.B. and Simpson, N.B. (2004), “Public education expenditure and growth”,
Journal of Development of Economics 73, 583-605.
Blankenau, F.B., Simpson, N.B. and Tomljanovich, M. (2007), “Public education expendi-
ture, taxation, and growth : Linking data to theory”, American Economic Review, Paper
and Proceeding 97(2), 393-397.
Blomquist, G.C. and Whitehead, J.C. (1998), “Resource Quality Information and Validity
of Willingness to Pay in Contingent Valuation”,Resource and Energy Economics, 20,
179 -196.
Boldrin, M., and Montes, A. (2005), “The intergenerational state education and pensions”,
Review of Economic Studies, 00, 1-14.
Bond, E., Trask, K. and Wang, P. (2003), “Factor accumulation and trade : dynamic compa-
rative advantage with endogenous physical and human capital”, International Economic
Review 44(3), 1041-1060.
Bond, EW., Wang, P. and Yip, CK. (1996), “A general two-sector model of endogenous
growth with human and physical capital : Balanced Growth and Transitional Dyna-
mics”, Journal of Economic Theory, 68, 149-173.
Bovenberg, A.L. and Smulders, S.A. (1996), “Transitional impacts of environmental po-
licy in an endogenous growth model”, International Economic Review 37, 861-893.
Brock, W. and Taylor, S. (2005), “Economic growth and the environment : a review
of theory and empirics”, In P. Aghion and S. Durlauf (eds.), Handbook of Economic
Growth, vol. I, part 2, 1749-1821. Amsterdam : Elsevier North-Holland.
Bräuninger, M. and Vidal, J.P. (2000), “Private versus public financing of education and
endogenous growth,” Journal of Population Economics, 13, 387-401.
Burstein, A.T., Nevec, J.C and Rebelo, S. (2004), “Investment Prices and Exchange Rates :
Some Basic Facts”, Journal of the European Economic Association, 302-309.
Cardak, B.A. (2004), “Education choice, neoclassical growth, and class structure”, Oxford
Economic Papers, 56(4), 643-666.
202
References
Chinn, M.D. (2000), “The Usual Suspects ? Productivity and Demand shocks and Asia-
pacific real exchange rates”, Review of International Economics, 8(1), 20-43.
Chiroleu-Assouline, M. (2001), “Le double dividende. Les approches théoriques ”, Revue
française d’économie, 16(2), 119-147.
Chiroleu-Assouline, M. and Fodha, M. (2006), “Double Dividend Hypothesis, Golden
Rule and Welfare Distribution”, Journal of Environmental Economics and Management,
51(3), 323-335.
Cohen, D. and Soto, M. (2007), “ Growth and human capital : good data, good results,”
Journal of Economic Growth, 12, 51-67.
Cremer, H., Pestieau, P., Thibault, E. and Vidal, J.P. (2006), “ Optimal tax and educa-
tion policy when agents differ in altruism and productivityy”, Annals of Economics and
Finance, 6(2), 209-228.
Cremer, H. and Pestieau, P. (2005) “ Intergenerational transfer of human capital and opti-
mal education policy”, Journal of Public Economic Theory, 8, 529-545.
Crumpler, H. and Grossman, P. (2008), “An Experimental Test of Warm Glow Giving”,
Journal of Public Economics, 92(5-6), 1011-1021.
Davin M., Gente K. and Nourry, C. (2012), “Social optimum in an OLG model with pa-
ternalistic altruism”, Economic Bulletin, 32(4) , 3417-3424.
De Fraja, G. (2002), “The design of optimal education policy”, Review of Economic Stu-
dies, 69, 437-466.
De La Croix, D. and Doepke, M. (2003), ”Inequality and growth : Why differential fertility
matters ?”, American Economic Review, 93, 1091-1113.
De la Croix, D. and Monfort, P. (2000) : “ Education Funding and Regional Convergence,”
Journal of Population Economics, 13, 403-424.
De la Fuente, A. and Doménech, R. (2006), “ Human capital in growth regressions : How
much differences does data quality make ?,” Journal of the European Economic Asso-
ciation, 4(1), 1-36.
203
References
Del Rey, E. and Lopez-Garcia, M-A. (2010), “On welfare criteria in Dynamic in an endo-
genous growth model”, Journal of Public Economic Theory, 14(6), 927–943.
Devereux, M. and Love, D. (1994), “The effect of factor taxation in a two-sector model of
endogenous growth”, Canadian Journal of Economics 275(3), 509-536.
Docquier, F., Paddison, O. and Pestieau, P. (2007), “Optimal Accumulation in an Endoge-
nous Growth Setting with Human Capital”, Journal of Economic Theory, 134, 361-378.
Dotsey, M. and Duarte, M. (2008), “Nontraded goods, market segmentation, and exchange
rates”, Journal of Monetary Economics, 55, 1129-1142.
Drugeon, J.P. (2004), “On Consumptions, Inputs and Outputs Substitutabilities and the
Evanescence of Optimal Cycles”, Journal of Difference Equations and Applications,
10(4), 473-487.
Dworkin, RM. (1977), “Taking rights seriously”, Harvard University Press, Cambridge.
European Commission (2008), “The Attitudes of European Citizens towards the Environ-
ment”, Eurobarometer 296, DG Environment.
Egger, H., Egger, P., Falkinger, F. and Grossmann, V. (2010), “The impact of capital mar-
ket integration on educational choice and the consequences for economic growth”, The
World Economy, 33, 1241–1268.
Erosa, A., Koreshkova, T. and Restuccia, D. (2012), “How important is human capital ? A
quantitative theory assessment of world income inequality”, The Review of Economic
Studies, 77, 1421-1449.
Foster, B.A. (1973), “Optimal capital accumulation in a polluted environment”, Southern
Economic Journal, 29(4), 544-547.
Gala, P. (2007), “Real exchange rate levels and economic development : theoretical analy-
sis and econometric evidence”,Cambridge Journal of Economics, 32(2), 273-288.
Galor, O. (1992), “A two-sector overlapping-generations model : a global characterization
of the dynamical system”, Econometrica, 60, 1351-1386.
204
References
Galor, O. (1992), “The choice of factor mobility in a dynamic world”, Journal of Popula-
tion Economics, 5, 135-144.
Galor, O. and Moav, O. (2004), “From physical to human capital accumulation : Inequality
and the process of development”, The Review of Economic Studies, vol.70, no.4, 1001-
1026.
Galor, O. and Mountford, A. (2008), “Trading population for productivity : theory and
evidence”, Review of Economic Studies, 75, 1143-1175.
Glomm, G. and Ravikumar, B. (1992), “Public versus private investment in human capital :
Endogenous growth and income inequality”, Journal of Political Economy, 100(4), 818-
833.
Glomm, G. and Ravikumar, B. (1997), “Productive Government Expenditures and Long
Run Growth”, Journal of Economic Dynamics and Control, 21, 183-204.
Glomm, G. and Ravikumar, B. (1998), “Flate-Rate Taxes, Government Spending on Edu-
cation, and Growth,” Review of Economic Dynamics, 1, 306-325.
Goulder, L.H. (1995), “Environmental Taxation and the Double Dividend : a Reader’s
Guide”, International Tax and Public Finance, 2, 157-183.
Gradus, R. and Smulders, S. (1993), “The Trade-off between Environmental Care and
Long-term Growth : Pollution in three Prototype Growth Models”, Journal of Econo-
mics, 58, 25-51.
Grimaud, A. and Tournemaine, F. (2007), “Why can an environmental policy tax promote
growth through the channel of education”, Ecological Economics, 62(1), 27-36.
Harsanyi, J.C. (1995) : “A theory of prudential values and a rule utilitarian theory of mo-
rality,” Social Choice and Welfare 12, 319-333.
Herrendorf, B. and Valentinyi, A. (2008) : “Measuring factor income shares at the sectoral
level”, Review of Economic Dynamics 11, 820-835.
Hseih, C. and Klenow, P. (2008), “Misallocation and manufacturing TFP in china and
india”, The Quarterly Journal of Economics vol. 124(4), 1403-1448.
205
References
Hu Y., Kemp, M. and Shimomura, K. (2009), “A two-country dynamic Heckscher-Ohlin
model with physical and human capital accumulation”, Economic Theory 41, 67-84.
Inklaar, R. and Timmer, M.P. (2012), “The Relative Price of Services”, Review of Income
and Wealth, DOI : 10.1111/roiw.12012
Ito T., Isard, P. and Symansky, S. (1999), “Economic growth and real exchange rate : an
overview of the Balassa-Samuelson hypothesis in Asia”, Changes in exchange rates
in rapidly developing countries : theory, practice, and policy issues, Univ. of Chicago
Press ISBN 0226386732, 109-128.
John, A. and Pecchenino, R. (1994), “An overlapping generations model of growth and the
environment”, Economic Journal, 104, 1393-1410.
Keane, M.P., and Wolpin, K.I. (2001), “The effect of parental transfers and borrowing
constraints on educational attainment”, International Economic Review, 42(4), 1051-
1103.
Kotchen, M. and Moore, M. (2008), “Conservation : From Voluntary Restraint to a Volun-
tary Price Premium”, Environmental and Resource Economics, 40(2), 195-215.
Krueger, A. and Lindahl, M. (2001), “Education for growth :Why and for whom ?”, Jour-
nal of Economic Literature, 39(4), 1101–1136.
Kutan, M.A. and Yigit, M.T. (2007), “European integration, productivity growth and real
convergence”, European Economic Review, 51,1370-1395.
Lombardo, G. and Ravenna, F. (2014), “Openness and optimal monetary policy”, Journal
of International Economics, In Press.
Lucas, R. (1988) : “On the mechanics of economic development”, Journal of Monetary
Economics, 22, 3-42.
Mankiw, G., Romer, D. and Weil, D. (1992), “ A contribution to the empirics of economic
growth”, Quarterly Journal of Economics, 107(2), 402–437.
Mariani, F., Pérez-Barahona, A. and Raffin, N. (2010), “Life expectancy and the environ-
ment”, Journal of Economic Dynamics and Control, 34(4), 798-815.
206
References
Menges, R., Schroeder, C., and Traub, S. (2005), “Altruism, warm glow and the
willingness-to-donate for green electricity : An artefactual field experiment”, Environ-
mental and Resource Economics, 31(4), 431-458.
Michel, P. and Vidal, J.P. (2000), “Economic integration and growth under intergeneratio-
nal financing of human-capital formation”, Journal of Economics, 72(3), 275-294.
Mincer, J. (1974), “Schooling, experience, and earnings”, National Bureau of Economic
Research, Book, Distributed by Columbia University Press.
Mincer, J. (1991), “Education and unemployment”, National Bureau of Economic Re-
search, 3838.
Mino, K., Nishimura, K., Shimomura, K. and Wang, P. (2008) : “Equilibrium dynamics
in discrete-time endogenous growth models with social constant returns”, Economic
Theory, 34, 1-23.
Mirowsky, J. and Ross, C.E. (1998), “Education, personal control, lifestyle and health : a
human capital hypothesis”,Research on Ageing, 20, 415-449.
Nelson, R. and Phelps, E. (1966), “Investment in humans, technological diffusion, and
economic growth”, American Economic Review, 56(1/2), 65-75.
OECD (2007), “Instrument Mixes for Environmental Policy”, OECD Publishing, Paris.
OECD (2008), “Promoting Sustainable Consumption - Good Practices in OECD Coun-
tries”, OECD Publishing, Paris.
OECD (2010), “Taxation, Innovation and the Environment”, OECD Publishing, Paris.
Ono, T. (2003a), “Environmental Tax Policy and Long Run Economic Growth”, Japanese
Economic Review, 54, 203-217.
Ono, T. (2003b), “Environmental Tax Policy in a Model of Growth Cycles”, Economic
Theory, 22, 141-168.
Oreopoulos, P., Page, M..E. and Stevens, A.H. (2006), “The Intergenerational effects of
compulsory schooling”, Journal of Labor Economics, 24(4), 729-760.
207
References
Pautrel, X. (2011), “Abatement Technology and the Environment-Growth Nexus with Edu-
cation”, LEMNA Working paper EA 4272.
Pautrel, X. (2012), “Environmental policy, Education and Growth : a Reappraisal when
Lifetime is Finite”, Macroeconomic Dynamics, 16(5), 611-685.
Pritchett, L. (2001), “Where Has All the Education Gone”, World Bank Economic Review,
15(3), 367-392.
Prieur, F. and Bréchet, T. (2013), “Can education be good for both growth and the envi-
ronment ?”, Macroeconomic Dynamics, 17(5), 1135-1157.
Rebelo, S. (1991), “Long-run policy analysis and long-run growth”, Journal of Political
Economy, 99(3), 500-521.
Ribar, D.C. and Wilhelm, M.O. (2002), “Altruistic and joy-of-giving motivations in chari-
table behavior”, Journal of Political Economy, 110(2), 425-457.
Rivera-Batiz, L.A. and Romer, P.M. (1991), “Economic integration and endogenous grow-
th”, The Quarterly Journal of Economics, 106(2), 531-555.
Rodrick, D. (2008), “The Real Exchange Rate and Economic Growth”, Brooking Papers
on Economic Activity, 39, 365-439.
Seegmuller, T. and Verchère, A. (2004), “Pollution as a Source of Endogenous Fluctua-
tions and Periodic Welfare Inequality in OLG Economies”, Economics Letters, 84, 363-
369.
Schob, R. (2005), “The double dividend hypothesis of environmental taxes : a survey”,
The International Yearbook of Environmental and Resource Economics 2005/2006, edi-
ted by H,Folmer and T.H, Tietenberg, 6, 223-280.
Schubert, K. (2009), “Pour la taxe carbone : la politique économique face à la menace
climatique”, Centre pour la recherche économique et ses applications, Opuscule no 18
- ISBN 978-2-7288-0431-3.
Schultz, T.W. (1961), “Investment in human capital”, The American Economic Review,
51(1), 1-17.
208
References
Schumacher, I. (2009), “The Dynamics of Environmentalism and the Environment”, Eco-
logical Economics, 68(11), 2842-2849.
Schumacher, I. and Zou, B. (2013), “Threshold Preferences and the Environment”, CREA
Discussion Paper, 2013-14.
Takahashi, H., Mashiyama, K. and Sakagami, T. (2012), “Does The Capital Intensity Mat-
ter ? Evidence From The Postwar Japanese Economy And Other Oecd Countries”, Ma-
croeconomic Dynamics, 16, 103-116.
Turnovsky, S.J. (2009), “Capital accumulation and economic growth in a small open eco-
nomy”, Cambridge University Press.
Uzawa, H. (1965) : “Optimal technical change in an aggregative model of economic
growth„” International Economic Review, 6,18-31.
Varvarigos, D. (2011), “Environmental Dynamics and the Links between Growth, Volati-
lity and Mortality,” Bulletin of Economic Research, 65(4), 314-331.
Venditti, A. (2005), “The two sector overlapping generations model : a simple formula-
tion”, Research in Economics, 59, 164-188.
Viaene, J.M. and Zilcha, I. (2002), “Public education under capital mobility”, Journal of
Economic Dynamics and Control, 26(2002), 2005–2036.
Wang, M, MO. Rieger and T. Hens (2011), “How time preferences differ : evidence from
45 countries”, Discussion Paper Series Number 182011, Department of Finance and
Management Science, Norwegian School of Economics.
Witzke, H.P. and Urfei, G. (2001), “Willingness to Pay for Environmental Protection in
Germany : Coping with the Regional Dimension”, Regional Studies, 35(3), 207-214.
Xepapadeas, A. (2005), “Economic Growth and the Environment”, In Handbook of Envi-
ronmental Economics, vol. III, 1219-1271, Elsevier Science.
Zhang, J. (1996), “Optimal public investment in education and endogenous growth”, Scan-
dinave Journal of Economics, 98, 387-404.
209
References
Zhang, J. (1999), “Environmental sustainability, nonlinear dynamics, and chaos”, Econo-
mic Theory, 14, 489-500.
Zuelta, H., and Young, A. (2012), “Labor’s shares in a model of induced innovation”,
Structural Change and Economic Dynamics, 24, 112-122.
210